framework,version,device,op_name,kernel_source,moe_dtype,num_tokens,hidden_size,inter_size,topk,num_experts,moe_tp_size,moe_ep_size,distribution,latency
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,1,1,balanced,0.356192626953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,1,1,balanced,0.6761286163330078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,1,1,balanced,0.6949523162841797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,1,1,balanced,0.20091136932373047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,1,1,balanced,0.6714419555664063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,1,1,balanced,0.6658380889892579
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,1,1,balanced,0.6764562988281251
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,1,1,balanced,0.6827986907958985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,1,1,balanced,0.6829388427734375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,1,1,balanced,0.6868415832519531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,1,1,balanced,0.6959347534179687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,1,1,balanced,0.7011436462402344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,1,1,balanced,0.7072128295898438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,1,1,balanced,0.7593138885498047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,1,1,balanced,0.7754208374023437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,1,1,balanced,0.7695142364501952
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,1,1,balanced,0.8059916687011718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,1,1,balanced,1.2415711975097656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,1,1,balanced,1.2731314849853514
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,1,1,balanced,1.93098876953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,1,1,balanced,2.5193951416015627
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,1,1,balanced,3.7708575439453127
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,1,1,balanced,4.838289184570312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,1,1,balanced,9.887124633789062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,1,1,balanced,7.266028442382813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,1,1,balanced,14.85195556640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,1,1,balanced,19.5646484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,1,1,balanced,0.37361854553222656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,1,1,balanced,0.3766239929199219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,1,1,balanced,0.2087411117553711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,1,1,balanced,0.3695916748046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,1,1,balanced,0.37968830108642576
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,1,1,balanced,0.38593536376953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,1,1,balanced,0.38944831848144534
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,1,1,balanced,0.37986495971679685
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,1,1,balanced,0.3943967819213867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,1,1,balanced,0.3793824005126953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,1,1,balanced,0.368732795715332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,1,1,balanced,0.3759846496582031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,1,1,balanced,0.13299136161804198
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,1,1,balanced,0.42263744354248045
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,1,1,balanced,0.43405120849609374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,1,1,balanced,0.44494655609130856
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,1,1,balanced,0.45714881896972653
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,1,1,balanced,0.659269790649414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,1,1,balanced,0.701553955078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,1,1,balanced,1.0126764678955078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,1,1,balanced,1.3501580810546874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,1,1,power_law_1.01,0.20382848739624024
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,1,1,power_law_1.01,0.3452467346191406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,1,1,power_law_1.01,0.4334758377075195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,1,1,power_law_1.01,0.6539635467529297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,1,1,power_law_1.01,0.6741446685791016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,1,1,balanced,2.070098571777344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,1,1,power_law_1.01,0.7057331085205079
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,1,1,power_law_1.01,0.6810752105712891
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,1,1,power_law_1.01,0.7558573150634766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,1,1,power_law_1.01,0.7757234954833985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,1,1,power_law_1.01,0.8139520263671874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,1,1,power_law_1.01,0.8502617645263673
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,1,1,power_law_1.01,0.7543865966796874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,1,1,power_law_1.01,0.9306412506103516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,1,1,power_law_1.01,0.981934051513672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,1,1,power_law_1.01,0.9970835113525391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,1,1,power_law_1.01,1.072157440185547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,1,1,balanced,2.6464199829101562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,1,1,power_law_1.01,1.1015122985839843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,1,1,power_law_1.01,1.4320991516113282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,1,1,balanced,4.011507263183594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,1,1,power_law_1.01,1.685047607421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,1,1,balanced,5.2139257812499995
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,1,1,power_law_1.01,2.9379418945312503
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,1,1,power_law_1.01,2.2950962829589843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,1,1,balanced,7.8899340820312505
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,1,1,power_law_1.01,4.085479125976563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,1,1,power_law_1.01,0.2058348846435547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,1,1,power_law_1.01,0.24455360412597654
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,1,1,power_law_1.01,0.373359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,1,1,power_law_1.01,0.1379411220550537
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,1,1,power_law_1.01,0.3701273727416992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,1,1,power_law_1.01,0.3763507080078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,1,1,power_law_1.01,0.37251007080078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,1,1,power_law_1.01,0.36814849853515624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,1,1,balanced,11.117197265625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,1,1,power_law_1.01,0.3703014373779297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,1,1,power_law_1.01,0.3905984115600586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,1,1,power_law_1.01,0.4381542587280274
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,1,1,power_law_1.01,0.48228607177734373
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,1,1,power_law_1.01,5.289946899414063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,1,1,power_law_1.01,0.485912971496582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,1,1,power_law_1.01,0.4937472152709961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,1,1,power_law_1.01,0.5171001434326172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,1,1,power_law_1.01,0.5549983978271484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,1,1,power_law_1.01,0.4180294418334961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,1,1,power_law_1.01,0.73857666015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,1,1,power_law_1.01,0.8933209228515626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,1,1,power_law_1.01,7.556802368164062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,1,1,power_law_1.01,1.226346206665039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,1,1,power_law_1.01,1.6039027404785156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,1,1,power_law_1.01,2.24106689453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,1,1,power_law_1.01,2.896460266113281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,1,1,power_law_1.2,0.20814464569091795
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,1,1,power_law_1.2,0.3317663955688476
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,1,1,power_law_1.2,0.3801996612548828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,1,1,power_law_1.01,10.450772705078126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,1,1,power_law_1.2,0.6550822448730469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,1,1,power_law_1.2,0.6450835418701172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,1,1,power_law_1.01,4.168690490722656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,1,1,power_law_1.2,0.7083033752441407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,1,1,power_law_1.2,0.6805107116699218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,1,1,power_law_1.2,0.7582067108154297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,1,1,power_law_1.2,0.7673177337646484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,1,1,power_law_1.2,0.7745919799804687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,1,1,power_law_1.2,0.822625274658203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,1,1,power_law_1.2,0.8477254486083984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,1,1,power_law_1.2,0.9314105224609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,1,1,power_law_1.2,0.9734841918945312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,1,1,power_law_1.2,1.0233177947998047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,1,1,power_law_1.01,5.4772711181640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,1,1,power_law_1.2,1.115134735107422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,1,1,power_law_1.2,1.1000685119628906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,1,1,power_law_1.2,1.443242950439453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,1,1,power_law_1.2,1.7167832946777346
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,1,1,power_law_1.2,2.28123779296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,1,1,power_law_1.01,15.58494873046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,1,1,power_law_1.2,2.843034973144531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,1,1,power_law_1.01,8.151279296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,1,1,power_law_1.2,4.116900329589844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,1,1,power_law_1.2,0.13461183547973632
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,1,1,power_law_1.2,0.19536256790161133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,1,1,power_law_1.2,0.21999359130859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,1,1,power_law_1.2,0.37296768188476564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,1,1,power_law_1.2,0.3545817565917969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,1,1,power_law_1.2,0.3770073699951172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,1,1,power_law_1.2,0.37270782470703123
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,1,1,power_law_1.01,19.9786962890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,1,1,power_law_1.2,5.233823852539063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,1,1,power_law_1.2,0.36781566619873046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,1,1,power_law_1.2,0.3811385726928711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,1,1,power_law_1.2,0.3931673431396484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,1,1,power_law_1.2,0.42109054565429693
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,1,1,power_law_1.2,0.43811199188232425
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,1,1,power_law_1.2,0.48010112762451174
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,1,1,power_law_1.2,0.48791168212890623
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,1,1,power_law_1.2,0.5044403076171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,1,1,power_law_1.2,0.5200166320800781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,1,1,power_law_1.2,0.5645254516601563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,1,1,power_law_1.2,0.7420012664794922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,1,1,power_law_1.2,0.8643526458740235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,1,1,power_law_1.2,1.2179878234863282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,1,1,power_law_1.01,11.230267333984376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,1,1,power_law_1.2,7.6771392822265625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,1,1,power_law_1.2,1.5034873962402344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,1,1,power_law_1.2,2.2300985717773436
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,1,1,power_law_1.2,10.57545166015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,1,1,power_law_1.2,2.8841702270507814
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,1,1,balanced,0.39962303161621093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,1,1,balanced,0.6229388809204102
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,1,1,balanced,1.166021728515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,1,1,power_law_1.2,4.2157830810546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,1,1,balanced,1.1754303741455079
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,1,1,balanced,1.804307861328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,1,1,balanced,1.8837983703613284
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,1,1,balanced,1.820828857421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,1,1,balanced,1.886066589355469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,1,1,balanced,1.9497702026367185
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,1,1,balanced,1.9678591918945312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,1,1,power_law_1.2,5.5186395263671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,1,1,balanced,1.9963002014160156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,1,1,balanced,2.0746246337890626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,1,1,balanced,2.1039993286132814
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,1,1,power_law_1.2,15.1143310546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,1,1,balanced,2.202414703369141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,1,1,balanced,2.881925048828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,1,1,balanced,2.7488507080078124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,1,1,balanced,2.5857760620117185
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,1,1,balanced,3.949393920898437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,1,1,balanced,4.837228698730469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,1,1,balanced,7.496555786132812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,1,1,power_law_1.2,8.118667602539062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,1,1,balanced,9.097011108398437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,1,1,balanced,14.222817382812499
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,1,1,power_law_1.2,20.3559716796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,1,1,balanced,0.19486528396606445
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,1,1,balanced,0.606173439025879
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,1,1,balanced,19.439436035156252
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,1,1,balanced,0.6524678039550781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,1,1,balanced,0.6511859130859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,1,1,balanced,0.32941951751708987
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,1,1,balanced,0.6589842987060547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,1,1,balanced,0.663154525756836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,1,1,balanced,0.6649759674072266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,1,1,balanced,0.6681452941894531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,1,1,balanced,0.6743622589111328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,1,1,balanced,0.6839289855957031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,1,1,power_law_1.2,11.266962890624999
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,1,1,balanced,0.6880691528320313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,1,1,balanced,0.7058393859863281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,1,1,balanced,0.7207154846191406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,1,1,balanced,0.6090995025634766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,1,1,balanced,0.734458236694336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,1,1,balanced,0.7657446289062501
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,1,1,balanced,1.1196339416503907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,1,1,balanced,1.1922144317626953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,1,1,balanced,1.7951910400390623
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,1,1,balanced,2.2871743774414064
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,1,1,balanced,3.501224365234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,1,1,balanced,4.772310485839844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,1,1,balanced,31.863601074218753
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,1,1,power_law_1.01,0.4058489608764648
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,1,1,balanced,6.610081176757812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,1,1,power_law_1.01,0.593908462524414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,1,1,power_law_1.01,0.7553510284423828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,1,1,power_law_1.01,1.1515392303466796
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,1,1,power_law_1.01,1.8106008911132812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,1,1,balanced,8.886171874999999
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,1,1,power_law_1.01,1.8294841003417968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,1,1,power_law_1.01,1.8865286254882812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,1,1,power_law_1.01,1.89053955078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,1,1,power_law_1.01,1.9196691894531248
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,1,1,power_law_1.01,2.019453430175781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,1,1,power_law_1.01,2.1126559448242186
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,1,1,power_law_1.01,2.1288517761230468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,1,1,balanced,13.32916259765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,1,1,power_law_1.01,2.4025920104980467
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,1,1,power_law_1.01,2.5450157165527343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,1,1,power_law_1.01,2.6006964111328124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,1,1,power_law_1.01,2.7141439819335935
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,1,1,power_law_1.01,3.299363098144531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,1,1,balanced,44.433046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,1,1,power_law_1.01,4.352773132324218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,1,1,balanced,17.69922607421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,1,1,power_law_1.01,5.401945190429688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,1,1,power_law_1.01,7.769936523437499
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,1,1,power_law_1.01,10.3332373046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,1,1,power_law_1.01,15.015214843749998
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,1,1,power_law_1.01,0.19807167053222657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,1,1,power_law_1.01,0.3103737640380859
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,1,1,power_law_1.01,0.382501106262207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,1,1,power_law_1.01,0.5971027374267578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,1,1,power_law_1.01,0.6520313262939453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,1,1,power_law_1.01,0.6573331451416016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,1,1,power_law_1.01,0.6636768341064453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,1,1,power_law_1.01,0.6549542236328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,1,1,power_law_1.01,19.5817626953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,1,1,power_law_1.01,0.6591903686523437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,1,1,power_law_1.01,0.6748121643066407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,1,1,power_law_1.01,0.678458251953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,1,1,balanced,67.6789892578125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,1,1,power_law_1.01,0.6690892791748047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,1,1,power_law_1.01,0.698344955444336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,1,1,power_law_1.01,0.7213107299804687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,1,1,power_law_1.01,0.744052505493164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,1,1,power_law_1.01,0.8262322998046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,1,1,power_law_1.01,0.9479782104492187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,1,1,power_law_1.01,1.2408428955078126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,1,1,power_law_1.01,1.5006643676757814
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,1,1,power_law_1.01,2.0533970642089843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,1,1,power_law_1.01,2.571556396484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,1,1,power_law_1.01,3.6712991333007814
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,1,1,power_law_1.01,30.78194091796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,1,1,power_law_1.01,4.686314392089844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,1,1,power_law_1.01,6.863953857421874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,1,1,power_law_1.2,0.40829631805419925
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,1,1,power_law_1.2,0.5660620880126953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,1,1,power_law_1.2,0.6603743743896484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,1,1,power_law_1.2,1.151964797973633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,1,1,balanced,92.6576171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,1,1,power_law_1.2,1.7129125976562498
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,1,1,power_law_1.01,9.528990478515626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,1,1,power_law_1.2,1.8253561401367187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,1,1,power_law_1.2,1.8992857360839843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,1,1,power_law_1.01,42.2610302734375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,1,1,power_law_1.2,1.8864083862304688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,1,1,power_law_1.2,1.9657632446289064
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,1,1,power_law_1.2,2.02935302734375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,1,1,power_law_1.2,2.118011474609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,1,1,power_law_1.2,2.1681677246093747
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,1,1,power_law_1.2,2.241961669921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,1,1,power_law_1.2,2.515500183105469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,1,1,power_law_1.2,2.610022888183594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,1,1,power_law_1.2,2.6514688110351563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,1,1,power_law_1.2,3.317831115722656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,1,1,power_law_1.01,14.063138427734376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,1,1,power_law_1.2,4.384563903808593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,1,1,power_law_1.2,5.405660400390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,1,1,power_law_1.2,7.6463513183593745
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,1,1,power_law_1.01,17.9287841796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,1,1,power_law_1.2,10.40273681640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,1,1,power_law_1.2,15.003471679687499
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,1,1,power_law_1.2,0.1967340850830078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,1,1,power_law_1.01,64.749638671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,1,1,power_law_1.2,0.2981043243408203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,1,1,power_law_1.2,0.34692222595214844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,1,1,power_law_1.2,0.5966527938842774
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,1,1,power_law_1.2,0.6285446548461915
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,1,1,power_law_1.2,0.6534982299804687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,1,1,power_law_1.2,0.6554374694824219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,1,1,power_law_1.2,0.649768295288086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,1,1,power_law_1.2,0.6641587066650391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,1,1,power_law_1.2,0.6645664215087891
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,1,1,power_law_1.2,0.6730150604248047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,1,1,power_law_1.2,0.6780953979492187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,1,1,power_law_1.2,19.560186767578124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,1,1,power_law_1.2,0.677606430053711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,1,1,power_law_1.2,0.7243846130371094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,1,1,power_law_1.2,0.7507334136962891
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,1,1,power_law_1.2,0.8640415954589844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,1,1,power_law_1.2,0.942008285522461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,1,1,power_law_1.2,1.1539199829101563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,1,1,power_law_1.2,1.4124461364746093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,1,1,power_law_1.2,2.040070343017578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,1,1,power_law_1.2,2.540166473388672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,1,1,power_law_1.2,3.6473684692382813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,1,1,power_law_1.2,31.1450732421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,1,1,power_law_1.2,4.694398803710937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,1,1,power_law_1.2,6.78904541015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,1,1,power_law_1.01,89.7016015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,1,1,balanced,0.20890495300292972
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,1,1,balanced,0.43650558471679685
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,1,1,power_law_1.2,9.4420849609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,1,1,power_law_1.2,42.815146484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,1,1,balanced,1.3363130187988281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,1,1,balanced,0.7052652740478516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,1,1,balanced,2.642694396972656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,1,1,balanced,5.2279443359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,1,1,balanced,5.236714477539063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,1,1,balanced,5.2401403808593745
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,1,1,power_law_1.2,13.918673095703124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,1,1,balanced,5.3404870605468755
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,1,1,balanced,5.582122192382813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,1,1,balanced,5.271387329101563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,1,1,balanced,5.318248901367188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,1,1,balanced,5.406760864257812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,1,1,balanced,5.378688354492188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,1,1,balanced,8.41844970703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,1,1,balanced,8.495847778320313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,1,1,power_law_1.2,18.327897949218748
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,1,1,power_law_1.2,65.7063330078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,1,1,balanced,8.622973022460936
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,1,1,balanced,9.05220947265625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,1,1,balanced,9.480096435546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,1,1,balanced,10.102090454101562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,1,1,balanced,10.526861572265625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,1,1,balanced,11.968704833984376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,1,1,balanced,12.783970947265626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,1,1,balanced,18.650599365234378
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,1,1,balanced,0.12565759658813475
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,1,1,balanced,22.98176513671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,1,1,balanced,0.2215519905090332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,1,1,balanced,1.3343994140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,1,1,balanced,0.4057491302490234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,1,1,power_law_1.2,90.40763671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,1,1,balanced,0.7083097839355469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,1,1,balanced,2.59325439453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,1,1,balanced,2.601956481933594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,1,1,balanced,2.609888000488281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,1,1,balanced,36.61667724609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,1,1,balanced,2.62013427734375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,1,1,balanced,2.6702175903320313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,1,1,balanced,2.695502624511719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,1,1,balanced,2.641350402832031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,1,1,balanced,2.6257913208007815
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,1,1,balanced,2.6553268432617188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,1,1,balanced,2.969732971191406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,1,1,balanced,46.4468212890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,1,1,balanced,3.295588989257813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,1,1,balanced,3.0063577270507813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,1,1,balanced,2.985907287597656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,1,1,balanced,3.090099182128906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,1,1,balanced,3.8037191772460934
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,1,1,balanced,3.148949890136719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,1,1,power_law_1.01,0.4378886413574219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,1,1,balanced,3.4148825073242186
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,1,1,power_law_1.01,0.7025708770751953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,1,1,balanced,6.787939453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,1,1,power_law_1.01,1.3362803649902344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,1,1,balanced,7.133668212890624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,1,1,power_law_1.01,2.0523143005371094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,1,1,balanced,9.836985473632812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,1,1,power_law_1.01,3.8054669189453123
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,1,1,power_law_1.01,4.099644775390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,1,1,balanced,12.069349365234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,1,1,balanced,4.106846618652344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,1,1,power_law_1.01,0.21337215423583986
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,1,1,power_law_1.01,4.446622009277344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,1,1,power_law_1.01,3.9958099365234374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,1,1,power_law_1.01,4.7989453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,1,1,power_law_1.01,4.9510324096679685
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,1,1,power_law_1.01,4.1412377929687505
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,1,1,power_law_1.01,5.502300415039063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,1,1,power_law_1.01,2.645294799804687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,1,1,power_law_1.01,7.407312622070313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,1,1,power_law_1.01,7.529926147460938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,1,1,power_law_1.01,8.005017700195314
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,1,1,power_law_1.01,8.704108276367188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,1,1,power_law_1.01,9.338817749023438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,1,1,power_law_1.01,10.280928955078124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,1,1,power_law_1.01,11.279721679687501
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,1,1,power_law_1.01,0.22058111190795898
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,1,1,power_law_1.01,15.558541259765624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,1,1,power_law_1.01,13.42523681640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,1,1,power_law_1.01,0.12509568214416505
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,1,1,power_law_1.01,1.0586214447021485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,1,1,power_law_1.01,20.63574951171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,1,1,power_law_1.01,1.990287322998047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,1,1,power_law_1.01,2.037292785644531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,1,1,power_law_1.01,1.3441510009765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,1,1,power_law_1.01,25.63309326171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,1,1,power_law_1.01,1.879853515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,1,1,power_law_1.01,2.0628326416015623
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,1,1,power_law_1.01,2.1662701416015624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,1,1,power_law_1.01,2.2467858886718752
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,1,1,power_law_1.01,0.4018751907348633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,1,1,power_law_1.01,2.259344024658203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,1,1,power_law_1.01,0.7375199890136719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,1,1,power_law_1.01,2.298756408691406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,1,1,power_law_1.01,2.6063494873046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,1,1,power_law_1.01,2.6385223388671877
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,1,1,power_law_1.01,2.7066400146484377
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,1,1,power_law_1.01,2.7924319458007814
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,1,1,power_law_1.01,2.99100341796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,1,1,power_law_1.01,3.4140750122070314
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,1,1,power_law_1.01,3.738241882324219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,1,1,power_law_1.01,4.4063085937499995
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,1,1,power_law_1.01,5.120028686523438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,1,1,power_law_1.01,36.10728759765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,1,1,power_law_1.2,0.7060991668701172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,1,1,power_law_1.01,6.360481567382813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,1,1,power_law_1.2,1.2361164855957032
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,1,1,power_law_1.01,7.974696655273438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,1,1,power_law_1.2,1.8525823974609374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,1,1,power_law_1.2,0.4337055969238281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,1,1,power_law_1.2,2.4832940673828126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,1,1,power_law_1.01,10.81913330078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,1,1,power_law_1.2,3.9006707763671877
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,1,1,power_law_1.2,0.22382144927978515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,1,1,power_law_1.2,4.019071350097656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,1,1,power_law_1.2,4.2512750244140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,1,1,power_law_1.2,3.710707092285156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,1,1,power_law_1.01,13.652877197265624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,1,1,power_law_1.01,46.2738525390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,1,1,power_law_1.2,4.37734130859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,1,1,power_law_1.2,4.852314147949219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,1,1,power_law_1.2,4.875984497070313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,1,1,power_law_1.2,6.068959350585938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,1,1,power_law_1.2,7.588563232421874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,1,1,power_law_1.2,7.8016204833984375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,1,1,power_law_1.2,8.164494628906251
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,1,1,power_law_1.2,8.9556787109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,1,1,power_law_1.2,9.493831176757812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,1,1,power_law_1.2,10.720894775390624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,1,1,power_law_1.2,11.39876708984375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,1,1,power_law_1.2,0.2206502342224121
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,1,1,power_law_1.2,14.0356494140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,1,1,power_law_1.2,0.12522560119628906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,1,1,power_law_1.2,0.6879436492919921
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,1,1,power_law_1.2,0.40220481872558594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,1,1,power_law_1.2,0.9705318450927735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,1,1,power_law_1.2,1.2671942138671874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,1,1,power_law_1.2,16.4234130859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,1,1,power_law_1.2,1.848851776123047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,1,1,power_law_1.2,1.9446694946289065
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,1,1,power_law_1.2,2.094962615966797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,1,1,power_law_1.2,2.140568389892578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,1,1,power_law_1.2,2.2047084045410155
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,1,1,power_law_1.2,21.194619140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,1,1,power_law_1.2,2.269319610595703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,1,1,power_law_1.2,2.3383763122558596
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,1,1,power_law_1.2,2.001578826904297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,1,1,power_law_1.2,2.6136691284179685
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,1,1,power_law_1.2,2.6847103881835936
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,1,1,power_law_1.2,2.7912493896484376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,1,1,power_law_1.2,3.0538842773437502
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,1,1,power_law_1.2,3.1083001708984375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,1,1,power_law_1.2,3.583326721191406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,1,1,power_law_1.2,26.415524902343748
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,1,1,power_law_1.2,3.794838256835938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,1,1,power_law_1.2,4.5397509765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,1,1,power_law_1.2,5.276865844726562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,1,1,power_law_1.2,6.659581298828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,1,1,power_law_1.2,7.922738647460937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,1,1,balanced,0.1805638313293457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,1,1,power_law_1.2,36.78380859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,1,1,power_law_1.2,11.05802490234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,1,1,balanced,0.6034284973144531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,1,1,balanced,0.3603839874267578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,1,1,power_law_1.2,13.908613281250002
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,1,1,balanced,4.503741455078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,1,1,balanced,1.1428979492187499
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,1,1,balanced,4.4925048828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,1,1,balanced,4.495359497070313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,1,1,balanced,2.2611244201660154
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,1,1,balanced,4.581799621582031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,1,1,balanced,4.561763916015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,1,1,balanced,4.672996520996094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,1,1,balanced,4.4647518920898435
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,1,1,balanced,5.171248168945313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,1,1,balanced,5.156912841796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,1,1,power_law_1.2,47.1539794921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,1,1,balanced,7.304555053710938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,1,1,balanced,7.3610888671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,1,1,balanced,7.480620117187501
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,1,1,balanced,7.805238647460937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,1,1,balanced,8.145384521484376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,1,1,balanced,8.583753051757814
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,1,1,balanced,8.906497192382812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,1,1,balanced,0.19260671615600586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,1,1,balanced,0.35078399658203124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,1,1,balanced,0.6332774353027344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,1,1,balanced,10.093824462890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,1,1,balanced,1.1499155426025391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,1,1,balanced,11.223463134765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,1,1,balanced,2.2460473632812503
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,1,1,balanced,2.2419885253906253
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,1,1,balanced,2.250469055175781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,1,1,balanced,0.1122316837310791
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,1,1,balanced,2.281949462890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,1,1,balanced,2.2632960510253906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,1,1,balanced,2.2777912902832034
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,1,1,balanced,2.2895353698730467
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,1,1,balanced,2.317229461669922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,1,1,balanced,2.3598950195312502
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,1,1,balanced,15.763848876953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,1,1,balanced,2.6947250366210938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,1,1,balanced,2.569359436035156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,1,1,balanced,2.6248455810546876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,1,1,balanced,2.6830535888671876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,1,1,balanced,2.8232192993164062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,1,1,balanced,19.585692138671874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,1,1,balanced,2.8918994140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,1,1,balanced,3.4427697753906252
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,1,1,balanced,3.4678604125976564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,1,1,balanced,3.737218017578125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,1,1,power_law_1.01,0.17749568939208984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,1,1,power_law_1.01,0.3644230270385742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,1,1,power_law_1.01,0.6034700775146484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,1,1,balanced,5.436690673828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,1,1,power_law_1.01,1.1416831970214845
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,1,1,power_law_1.01,1.7555474853515627
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,1,1,balanced,5.918974609375001
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,1,1,power_law_1.01,3.22894287109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,1,1,balanced,8.522981567382812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,1,1,power_law_1.01,3.4301867675781255
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,1,1,power_law_1.01,3.528081359863281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,1,1,balanced,31.44089599609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,1,1,balanced,10.71358642578125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,1,1,power_law_1.01,2.260777587890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,1,1,power_law_1.01,3.604244384765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,1,1,power_law_1.01,3.7507553100585938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,1,1,power_law_1.01,4.2388742065429685
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,1,1,power_law_1.01,4.084371337890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,1,1,power_law_1.01,4.582515869140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,1,1,power_law_1.01,6.407444458007812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,1,1,power_law_1.01,6.502595825195312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,1,1,power_law_1.01,6.885740356445313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,1,1,balanced,40.31160888671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,1,1,power_law_1.01,7.433208618164062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,1,1,power_law_1.01,7.903252563476562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,1,1,power_law_1.01,8.609392700195311
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,1,1,power_law_1.01,9.439736328124999
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,1,1,power_law_1.01,0.11324864387512208
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,1,1,power_law_1.01,0.3505945587158203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,1,1,power_law_1.01,0.6349504089355469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,1,1,power_law_1.01,11.124776611328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,1,1,power_law_1.01,0.913376007080078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,1,1,power_law_1.01,1.1607129669189453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,1,1,power_law_1.01,0.19267391204833983
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,1,1,power_law_1.01,13.118353271484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,1,1,power_law_1.01,1.6201702880859377
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,1,1,power_law_1.01,1.7204774475097657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,1,1,power_law_1.01,1.7545753479003907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,1,1,power_law_1.01,1.7873989868164064
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,1,1,power_law_1.01,1.8634176635742186
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,1,1,power_law_1.01,1.9328608703613281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,1,1,power_law_1.01,1.9558892822265626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,1,1,power_law_1.01,1.9951148986816407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,1,1,power_law_1.01,17.37813720703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,1,1,power_law_1.01,2.202686767578125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,1,1,power_law_1.01,2.277238464355469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,1,1,power_law_1.01,2.3468199157714844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,1,1,power_law_1.01,2.43516357421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,1,1,power_law_1.01,2.6290093994140626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,1,1,power_law_1.01,2.8252517700195314
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,1,1,power_law_1.01,21.83165283203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,1,1,power_law_1.01,3.179598693847656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,1,1,power_law_1.01,3.730638122558594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,1,1,power_law_1.01,4.283123779296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,1,1,power_law_1.2,0.3634239959716797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,1,1,power_law_1.01,5.605298461914062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,1,1,power_law_1.2,0.6048012924194336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,1,1,power_law_1.2,1.0606732940673829
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,1,1,power_law_1.2,1.5849049377441407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,1,1,power_law_1.01,6.7531018066406245
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,1,1,power_law_1.2,2.1147853088378907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,1,1,power_law_1.01,30.38065673828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,1,1,power_law_1.2,0.17584896087646484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,1,1,power_law_1.2,3.176561279296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,1,1,power_law_1.01,9.548039550781251
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,1,1,power_law_1.2,3.335382995605469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,1,1,power_law_1.2,3.4308779907226565
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,1,1,power_law_1.2,3.5906610107421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,1,1,power_law_1.2,3.7454925537109376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,1,1,power_law_1.01,11.974875488281251
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,1,1,power_law_1.2,4.152972106933594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,1,1,power_law_1.2,4.225650634765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,1,1,power_law_1.2,4.845972595214844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,1,1,power_law_1.01,39.394677734375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,1,1,power_law_1.2,6.547656860351563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,1,1,power_law_1.2,6.731217651367187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,1,1,power_law_1.2,7.00989013671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,1,1,power_law_1.2,7.610105590820313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,1,1,power_law_1.2,8.0483642578125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,1,1,power_law_1.2,9.01277587890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,1,1,power_law_1.2,9.699093627929688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,1,1,power_law_1.2,0.19400831222534182
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,1,1,power_law_1.2,0.34990398406982426
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,1,1,power_law_1.2,0.11263487815856935
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,1,1,power_law_1.2,0.590827522277832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,1,1,power_law_1.2,0.8370719909667969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,1,1,power_law_1.2,11.760572509765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,1,1,power_law_1.2,1.0897030639648437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,1,1,power_law_1.2,1.593429718017578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,1,1,power_law_1.2,1.68272705078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,1,1,power_law_1.2,13.864522705078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,1,1,power_law_1.2,1.719959716796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,1,1,power_law_1.2,1.7918054199218751
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,1,1,power_law_1.2,1.8482623291015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,1,1,power_law_1.2,1.909629364013672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,1,1,power_law_1.2,1.9552108764648437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,1,1,power_law_1.2,2.021470031738281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,1,1,power_law_1.2,17.934342041015626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,1,1,power_law_1.2,2.285655059814453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,1,1,power_law_1.2,2.392878112792969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,1,1,power_law_1.2,2.3618789672851563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,1,1,power_law_1.2,2.61600830078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,1,1,power_law_1.2,2.6791506958007814
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,1,1,power_law_1.2,3.0829428100585936
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,1,1,power_law_1.2,3.2359878540039064
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,1,1,power_law_1.2,22.502568359374997
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,1,1,power_law_1.2,3.9478823852539064
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,1,1,power_law_1.2,4.600765991210937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,1,1,power_law_1.2,5.7903509521484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,1,1,balanced,0.06335807800292968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,1,1,power_law_1.2,7.083197631835938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,1,1,balanced,0.1038815975189209
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,1,1,balanced,0.0432908821105957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,1,1,power_law_1.2,30.921630859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,1,1,power_law_1.2,9.756037597656249
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,1,1,balanced,0.16695552825927734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,1,1,balanced,0.2990483283996582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,1,1,balanced,0.32111488342285155
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,1,1,power_law_1.2,12.19921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,1,1,balanced,0.30898624420166015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,1,1,balanced,0.34417793273925784
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,1,1,balanced,0.3084761619567871
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,1,1,balanced,0.3328736114501953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,1,1,balanced,0.31143999099731445
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,1,1,balanced,0.31326271057128907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,1,1,balanced,0.33901950836181644
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,1,1,power_law_1.2,40.289218749999996
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,1,1,balanced,0.3458099365234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,1,1,balanced,0.3586361694335937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,1,1,balanced,0.336126708984375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,1,1,balanced,0.3398252868652344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,1,1,balanced,0.3423436737060547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,1,1,balanced,0.3987839889526367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,1,1,balanced,0.42647937774658207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,1,1,balanced,0.03325952053070068
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,1,1,balanced,0.046638078689575195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,1,1,balanced,0.07079487800598144
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,1,1,balanced,0.10860032081604003
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,1,1,balanced,0.18011775970458985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,1,1,balanced,0.18093696594238282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,1,1,balanced,0.1858464050292969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,1,1,balanced,0.1872755241394043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,1,1,balanced,2.6452487182617186
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,1,1,balanced,0.18870271682739256
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,1,1,balanced,0.1916524887084961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,1,1,balanced,0.47942527770996096
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,1,1,balanced,0.19163455963134765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,1,1,balanced,0.21687936782836914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,1,1,balanced,0.2166873550415039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,1,1,balanced,0.21930496215820314
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,1,1,balanced,0.22359424591064453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,1,1,balanced,0.6561203002929688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,1,1,balanced,0.22540096282958982
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,1,1,balanced,0.23023296356201173
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,1,1,balanced,0.2429612731933594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,1,1,balanced,0.261333122253418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,1,1,balanced,0.2779353523254394
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,1,1,balanced,0.303956470489502
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,1,1,balanced,0.4444575881958007
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,1,1,balanced,0.49282497406005865
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,1,1,balanced,0.7044595336914062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,1,1,balanced,0.924133758544922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,1,1,balanced,0.6992447662353516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,1,1,power_law_1.01,0.09563520431518555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,1,1,balanced,1.4073484802246095
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,1,1,balanced,1.051095657348633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,1,1,power_law_1.01,0.17460544586181642
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,1,1,balanced,1.818995819091797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,1,1,balanced,1.34603515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,1,1,power_law_1.01,0.2885356712341308
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,1,1,power_law_1.01,0.058835840225219725
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,1,1,power_law_1.01,0.29790271759033204
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,1,1,balanced,2.0408543395996093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,1,1,power_law_1.01,0.2921459197998047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,1,1,power_law_1.01,0.3190745544433594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,1,1,power_law_1.01,0.3277356719970703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,1,1,power_law_1.01,0.13368063926696777
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,1,1,power_law_1.01,0.3514956665039063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,1,1,power_law_1.01,0.36631423950195313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,1,1,power_law_1.01,0.3517849731445312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,1,1,power_law_1.01,0.2542905616760254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,1,1,power_law_1.01,0.42254913330078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,1,1,power_law_1.01,0.2766073608398437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,1,1,power_law_1.01,0.0412716817855835
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,1,1,power_law_1.01,0.31010175704956056
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,1,1,power_law_1.01,0.38180736541748045
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,1,1,power_law_1.01,1.5040153503417968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,1,1,power_law_1.01,2.2163352966308594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,1,1,power_law_1.01,0.03132607936859131
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,1,1,power_law_1.01,2.813359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,1,1,power_law_1.01,0.046695680618286134
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,1,1,power_law_1.01,0.07062655925750733
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,1,1,power_law_1.01,0.09337792396545411
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,1,1,power_law_1.01,0.38757183074951174
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,1,1,power_law_1.01,0.47551422119140624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,1,1,power_law_1.01,0.11124608039855957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,1,1,power_law_1.01,0.1555609607696533
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,1,1,power_law_1.01,0.16862464904785157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,1,1,power_law_1.01,0.5684812927246095
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,1,1,power_law_1.01,0.1781337547302246
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,1,1,power_law_1.01,0.17837568283081057
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,1,1,power_law_1.01,0.713119354248047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,1,1,power_law_1.01,0.21093568801879883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,1,1,power_law_1.01,0.21652799606323242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,1,1,power_law_1.01,0.22558975219726562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,1,1,power_law_1.01,0.2262323188781738
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,1,1,power_law_1.01,0.23736383438110353
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,1,1,power_law_1.01,0.17266176223754884
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,1,1,power_law_1.01,0.2534310340881348
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,1,1,power_law_1.01,0.18427583694458008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,1,1,power_law_1.01,0.31445440292358395
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,1,1,power_law_1.01,0.2110233688354492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,1,1,power_law_1.01,0.3695609664916992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,1,1,power_law_1.01,0.8468326568603516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,1,1,power_law_1.01,0.46832767486572263
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,1,1,power_law_1.01,1.180688018798828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,1,1,power_law_1.01,0.5656447982788085
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,1,1,power_law_1.01,0.2727673530578613
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,1,1,power_law_1.2,0.04095808029174804
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,1,1,power_law_1.01,0.7979033660888672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,1,1,power_law_1.01,1.023501434326172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,1,1,power_law_1.01,1.4500186157226562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,1,1,power_law_1.01,1.911285705566406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,1,1,power_law_1.2,0.3004761505126953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,1,1,power_law_1.2,0.05890816211700439
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,1,1,power_law_1.2,0.09079232215881347
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,1,1,power_law_1.2,0.12517888069152833
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,1,1,power_law_1.2,0.2569068717956543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,1,1,power_law_1.2,0.3204985427856445
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,1,1,power_law_1.2,0.33323776245117187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,1,1,power_law_1.2,0.35253311157226563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,1,1,power_law_1.2,0.36814399719238283
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,1,1,power_law_1.2,0.16096960067749025
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,1,1,power_law_1.2,0.3874630355834961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,1,1,power_law_1.2,0.2788915252685547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,1,1,power_law_1.2,0.43287807464599604
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,1,1,power_law_1.2,0.2914713668823242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,1,1,power_law_1.2,0.3531039810180664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,1,1,power_law_1.2,0.5765644836425781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,1,1,power_law_1.2,0.29885824203491207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,1,1,power_law_1.2,0.7332793426513672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,1,1,power_law_1.2,0.3090752029418945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,1,1,power_law_1.2,0.9047135925292968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,1,1,power_law_1.2,0.04765376091003418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,1,1,power_law_1.2,0.0691097593307495
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,1,1,power_law_1.2,0.08639871597290039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,1,1,power_law_1.2,0.10410367965698242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,1,1,power_law_1.2,0.1579423999786377
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,1,1,power_law_1.2,0.38404159545898436
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,1,1,power_law_1.2,0.16998016357421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,1,1,power_law_1.2,0.17761407852172853
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,1,1,power_law_1.2,0.18042688369750975
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,1,1,power_law_1.2,0.18061119079589844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,1,1,power_law_1.2,1.2360940551757813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,1,1,power_law_1.2,0.18395776748657228
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,1,1,power_law_1.2,0.21107711791992187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,1,1,power_law_1.2,0.21518592834472655
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,1,1,power_law_1.2,0.22003583908081054
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,1,1,power_law_1.2,0.2258086395263672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,1,1,power_law_1.2,0.2305049514770508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,1,1,power_law_1.2,0.2393971252441406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,1,1,power_law_1.2,0.25373311996459963
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,1,1,power_law_1.2,0.2786092758178711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,1,1,power_law_1.2,0.3224844741821289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,1,1,power_law_1.2,0.3727743911743164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,1,1,power_law_1.2,0.47945598602294925
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,1,1,power_law_1.2,0.03009664058685303
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,1,1,power_law_1.2,0.493493766784668
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,1,1,power_law_1.2,0.5817772674560546
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,1,1,power_law_1.2,1.5670713806152343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,1,1,balanced,0.09603648185729981
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,1,1,balanced,0.18006975173950196
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,1,1,balanced,0.3116076850891113
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,1,1,balanced,0.5902163314819335
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,1,1,power_law_1.2,0.8045164489746094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,1,1,balanced,1.1360352325439453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,1,1,balanced,1.1442527770996094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,1,1,balanced,1.1512838745117189
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,1,1,balanced,1.1619391632080078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,1,1,balanced,1.1633209228515624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,1,1,balanced,1.170508804321289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,1,1,balanced,1.1810713958740233
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,1,1,balanced,1.762255401611328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,1,1,balanced,1.7770758056640623
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,1,1,balanced,1.8187295532226564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,1,1,power_law_1.2,1.0277881622314453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,1,1,balanced,1.8748921203613282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,1,1,balanced,1.9012486267089845
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,1,1,balanced,1.9860914611816405
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,1,1,balanced,2.1273849487304686
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,1,1,balanced,2.2172230529785155
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,1,1,balanced,2.559991607666016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,1,1,balanced,2.7514309692382812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,1,1,power_law_1.2,2.214581756591797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,1,1,balanced,3.9098770141601564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,1,1,balanced,4.961216125488281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,1,1,balanced,0.07057472229003905
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,1,1,balanced,0.1133024024963379
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,1,1,balanced,0.17804288864135742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,1,1,balanced,0.31236671447753905
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,1,1,balanced,0.5847603225708008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,1,1,balanced,7.408151245117187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,1,1,balanced,0.5892953491210937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,1,1,balanced,0.5989651107788087
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,1,1,balanced,0.6000934219360351
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,1,1,power_law_1.2,1.4478501892089843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,1,1,balanced,0.6038111877441407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,1,1,balanced,0.6094348907470704
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,1,1,balanced,0.614724464416504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,1,1,balanced,0.6799417877197266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,1,1,balanced,0.6714195251464844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,1,1,balanced,9.25641845703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,1,1,balanced,0.6819647979736329
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,1,1,balanced,0.7004793548583985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,1,1,balanced,0.7040831756591797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,1,1,balanced,0.7180019378662109
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,1,1,balanced,0.7422444915771484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,1,1,balanced,0.78272705078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,1,1,balanced,0.8639398193359374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,1,1,balanced,0.9054227447509765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,1,1,balanced,1.3492448425292969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,1,1,balanced,1.5754841613769532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,1,1,balanced,2.2895948791503904
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,1,1,balanced,2.9275296020507815
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,1,1,power_law_1.01,0.19007232666015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,1,1,power_law_1.01,0.3114067268371582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,1,1,balanced,14.140283203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,1,1,power_law_1.01,0.46227455139160156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,1,1,balanced,4.629927062988282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,1,1,power_law_1.01,0.09503744125366212
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,1,1,power_law_1.01,0.9463225555419921
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,1,1,power_law_1.01,1.0152063751220703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,1,1,power_law_1.01,1.0395673370361327
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,1,1,balanced,5.7039337158203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,1,1,power_law_1.01,1.0853702545166015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,1,1,power_law_1.01,1.0876863861083985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,1,1,power_law_1.01,1.176253433227539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,1,1,power_law_1.01,1.7059353637695314
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,1,1,power_law_1.01,1.7359475708007814
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,1,1,power_law_1.01,1.7901945495605467
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,1,1,power_law_1.01,1.883655700683594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,1,1,power_law_1.2,2.825500793457031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,1,1,balanced,19.349125976562497
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,1,1,power_law_1.01,1.8914399719238282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,1,1,power_law_1.01,1.981908416748047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,1,1,power_law_1.01,0.6003635025024414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,1,1,power_law_1.01,2.1266770935058594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,1,1,power_law_1.01,2.304509429931641
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,1,1,power_law_1.01,2.730732116699219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,1,1,power_law_1.01,3.2757678222656255
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,1,1,power_law_1.01,0.06539775848388671
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,1,1,power_law_1.01,0.10547007560729979
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,1,1,power_law_1.2,1.9259103393554688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,1,1,power_law_1.01,4.3304772949218755
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,1,1,power_law_1.01,0.17312639236450195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,1,1,power_law_1.01,0.2525472068786621
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,1,1,power_law_1.01,0.32208129882812503
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,1,1,power_law_1.01,0.4895040130615234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,1,1,power_law_1.01,5.457632446289063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,1,1,power_law_1.01,0.5275398254394531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,1,1,power_law_1.01,0.5426208114624023
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,1,1,power_law_1.01,0.560252799987793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,1,1,power_law_1.01,0.5817401504516602
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,1,1,power_law_1.01,0.5636556625366211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,1,1,power_law_1.01,0.6492115020751953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,1,1,power_law_1.01,0.647448959350586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,1,1,power_law_1.01,0.6724972534179687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,1,1,power_law_1.01,0.6930079650878906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,1,1,power_law_1.01,0.6976435089111328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,1,1,power_law_1.01,7.5494537353515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,1,1,power_law_1.01,0.7257183837890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,1,1,power_law_1.01,0.7661958312988281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,1,1,power_law_1.01,0.8374015808105468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,1,1,power_law_1.01,1.1506822204589844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,1,1,power_law_1.01,1.1916429138183593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,1,1,power_law_1.01,1.515667266845703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,1,1,power_law_1.01,9.784992065429687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,1,1,power_law_1.2,0.09800895690917968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,1,1,power_law_1.01,1.85025146484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,1,1,power_law_1.2,0.18005695343017578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,1,1,power_law_1.2,0.30430208206176756
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,1,1,power_law_1.2,0.42844799041748044
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,1,1,power_law_1.01,2.526961212158203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,1,1,power_law_1.2,0.5606131362915039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,1,1,power_law_1.01,3.1907064819335935
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,1,1,power_law_1.2,0.9542816162109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,1,1,power_law_1.2,1.0284646606445311
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,1,1,power_law_1.2,1.069019546508789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,1,1,power_law_1.2,1.1053913879394532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,1,1,power_law_1.2,1.1720416259765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,1,1,power_law_1.01,14.16471435546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,1,1,power_law_1.01,4.483309936523438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,1,1,power_law_1.2,1.7277253723144532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,1,1,power_law_1.2,1.7483091735839842
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,1,1,power_law_1.2,1.1186656188964843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,1,1,power_law_1.01,6.2079400634765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,1,1,power_law_1.2,1.8318663024902342
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,1,1,power_law_1.2,1.8927020263671874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,1,1,power_law_1.2,1.9197305297851563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,1,1,power_law_1.2,1.9993356323242188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,1,1,power_law_1.2,2.1772306823730467
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,1,1,power_law_1.2,2.352058868408203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,1,1,power_law_1.2,2.859307556152344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,1,1,power_law_1.2,3.3157849121093745
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,1,1,power_law_1.01,18.560687255859374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,1,1,power_law_1.2,0.11494784355163574
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,1,1,power_law_1.2,0.17003391265869142
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,1,1,power_law_1.2,4.362320556640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,1,1,power_law_1.2,0.23170623779296876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,1,1,power_law_1.2,0.3046086311340332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,1,1,power_law_1.2,0.4969055938720703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,1,1,power_law_1.2,0.5340607833862305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,1,1,power_law_1.2,0.5523654556274413
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,1,1,power_law_1.2,5.4522943115234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,1,1,power_law_1.2,0.5679142379760742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,1,1,power_law_1.2,0.5669760131835938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,1,1,power_law_1.2,0.580294418334961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,1,1,power_law_1.2,0.6572627258300782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,1,1,power_law_1.2,0.6579596710205078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,1,1,power_law_1.2,0.67271484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,1,1,power_law_1.2,0.6975251007080078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,1,1,power_law_1.2,0.7002681732177735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,1,1,power_law_1.2,0.7280223846435547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,1,1,power_law_1.2,0.7887602996826172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,1,1,power_law_1.2,0.874988784790039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,1,1,power_law_1.2,1.10465087890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,1,1,power_law_1.2,7.582642822265624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,1,1,power_law_1.2,1.101146240234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,1,1,power_law_1.2,0.06836351871490479
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,1,1,power_law_1.2,1.4909913635253906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,1,1,power_law_1.2,1.86433349609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,1,1,balanced,0.20323328018188475
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,1,1,power_law_1.2,2.5523143005371094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,1,1,balanced,0.3932051086425781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,1,1,power_law_1.2,9.996180419921874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,1,1,balanced,0.756671371459961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,1,1,balanced,1.457841339111328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,1,1,balanced,2.8171398925781252
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,1,1,balanced,3.519070129394531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,1,1,power_law_1.2,3.224813537597656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,1,1,balanced,3.5364184570312505
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,1,1,balanced,3.538477478027344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,1,1,balanced,3.625106506347656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,1,1,balanced,3.5924673461914063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,1,1,balanced,3.933740234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,1,1,balanced,3.8082757568359376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,1,1,balanced,5.765223388671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,1,1,balanced,5.859388427734375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,1,1,balanced,5.946580200195312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,1,1,balanced,6.098436279296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,1,1,power_law_1.2,14.45016845703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,1,1,power_law_1.2,4.506119079589844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,1,1,balanced,6.27884765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,1,1,balanced,6.566221923828126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,1,1,balanced,6.774270019531249
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,1,1,balanced,7.707802734374999
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,1,1,balanced,8.0992724609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,1,1,balanced,11.182705078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,1,1,balanced,0.14742976188659668
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,1,1,balanced,0.23790016174316406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,1,1,balanced,0.42395839691162107
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,1,1,balanced,0.7475193786621094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,1,1,balanced,14.193778076171876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,1,1,balanced,1.4166233825683594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,1,1,balanced,1.7880645751953126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,1,1,balanced,1.7763404846191406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,1,1,balanced,1.7962188720703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,1,1,power_law_1.2,5.834835205078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,1,1,power_law_1.2,18.833173828125002
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,1,1,balanced,19.444859619140622
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,1,1,balanced,1.7941139221191407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,1,1,balanced,1.7975917053222656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,1,1,balanced,1.8333509826660155
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,1,1,balanced,1.8291020202636719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,1,1,balanced,1.990113220214844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,1,1,balanced,2.0914137268066404
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,1,1,balanced,2.0806547546386716
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,1,1,balanced,2.0633818054199216
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,1,1,balanced,2.1430239868164063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,1,1,balanced,2.1549356079101565
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,1,1,balanced,2.259021453857422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,1,1,balanced,2.3696858215332033
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,1,1,balanced,2.5235078430175784
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,1,1,balanced,27.154799804687503
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,1,1,balanced,3.9638311767578123
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,1,1,power_law_1.01,0.2053772735595703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,1,1,balanced,4.48586181640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,1,1,power_law_1.01,0.3927686309814453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,1,1,power_law_1.01,0.75657470703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,1,1,power_law_1.01,1.2242675018310547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,1,1,balanced,6.00345458984375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,1,1,power_law_1.01,1.6775993347167968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,1,1,balanced,7.773438720703124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,1,1,power_law_1.01,2.7497940063476562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,1,1,power_law_1.01,2.96503173828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,1,1,power_law_1.01,3.030798034667969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,1,1,balanced,9.96703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,1,1,power_law_1.01,3.1444308471679685
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,1,1,power_law_1.01,3.271733703613281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,1,1,power_law_1.01,3.5250970458984376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,1,1,power_law_1.01,3.9503533935546877
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,1,1,balanced,13.63664306640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,1,1,power_law_1.01,5.345574951171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,1,1,power_law_1.01,5.640089111328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,1,1,balanced,40.332790527343754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,1,1,power_law_1.01,5.783692626953124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,1,1,power_law_1.01,5.833801879882812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,1,1,power_law_1.01,6.169025268554687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,1,1,power_law_1.01,6.820218505859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,1,1,power_law_1.01,7.135721435546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,1,1,power_law_1.01,8.000984497070313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,1,1,power_law_1.01,9.324783325195312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,1,1,power_law_1.01,0.14597503662109373
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,1,1,power_law_1.01,0.23774784088134765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,1,1,power_law_1.01,0.4217484664916992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,1,1,power_law_1.01,11.7791845703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,1,1,power_law_1.01,0.6462675476074219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,1,1,power_law_1.01,0.8618156433105468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,1,1,power_law_1.01,1.3754905700683593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,1,1,power_law_1.01,1.4706809997558594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,1,1,balanced,55.530063476562496
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,1,1,power_law_1.01,14.412987060546874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,1,1,power_law_1.01,1.5234757995605468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,1,1,power_law_1.01,1.5812384033203126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,1,1,power_law_1.01,1.6080519104003905
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,1,1,power_law_1.01,1.6462611389160156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,1,1,power_law_1.01,1.6771717834472657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,1,1,power_law_1.01,1.8489759826660155
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,1,1,power_law_1.01,1.9417234802246093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,1,1,power_law_1.01,1.9793312072753906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,1,1,power_law_1.01,1.9119308471679688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,1,1,power_law_1.01,19.918382568359373
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,1,1,power_law_1.01,2.0199731445312503
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,1,1,power_law_1.01,2.1915802001953124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,1,1,power_law_1.01,2.3184332275390624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,1,1,power_law_1.01,2.5644454956054688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,1,1,power_law_1.01,3.003801574707031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,1,1,power_law_1.01,3.7951986694335935
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,1,1,power_law_1.01,4.5422970581054685
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,1,1,power_law_1.2,0.20800832748413084
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,1,1,power_law_1.01,6.1666369628906255
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,1,1,power_law_1.01,25.42312744140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,1,1,power_law_1.2,0.7291238403320313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,1,1,power_law_1.2,0.39140350341796876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,1,1,power_law_1.2,1.0828562927246093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,1,1,power_law_1.01,7.471524047851562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,1,1,power_law_1.2,1.5382028198242188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,1,1,power_law_1.2,2.7029254150390623
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,1,1,power_law_1.2,2.894217529296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,1,1,power_law_1.2,3.0727578735351564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,1,1,power_law_1.2,3.2365048217773436
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,1,1,power_law_1.01,10.573397216796874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,1,1,power_law_1.2,3.3640948486328126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,1,1,power_law_1.2,3.6922079467773434
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,1,1,power_law_1.2,3.854352722167969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,1,1,power_law_1.01,13.907933349609374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,1,1,power_law_1.2,5.572251586914062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,1,1,power_law_1.01,36.46162109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,1,1,power_law_1.2,5.6966967773437505
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,1,1,power_law_1.2,5.988712158203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,1,1,power_law_1.2,6.123826293945312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,1,1,power_law_1.2,6.344465942382813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,1,1,power_law_1.2,6.798917846679688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,1,1,power_law_1.2,7.200430908203126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,1,1,power_law_1.2,8.370430297851563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,1,1,power_law_1.2,9.55436279296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,1,1,power_law_1.01,47.604755859375004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,1,1,power_law_1.2,0.14544447898864746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,1,1,power_law_1.2,0.23861312866210938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,1,1,power_law_1.2,0.41229503631591796
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,1,1,power_law_1.2,0.5869625473022462
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,1,1,power_law_1.2,12.22371826171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,1,1,power_law_1.2,0.7957292938232422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,1,1,power_law_1.2,1.3569868469238282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,1,1,power_law_1.2,1.4565292358398438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,1,1,power_law_1.2,1.5410258483886718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,1,1,power_law_1.2,14.774111328124999
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,1,1,power_law_1.2,1.5959942626953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,1,1,power_law_1.2,1.6493458557128906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,1,1,power_law_1.2,1.6877638244628905
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,1,1,power_law_1.2,1.7150054931640626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,1,1,power_law_1.2,1.8988371276855467
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,1,1,power_law_1.2,1.9673951721191407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,1,1,power_law_1.2,1.9801805114746096
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,1,1,power_law_1.2,2.034246368408203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,1,1,power_law_1.2,2.0901887512207034
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,1,1,power_law_1.2,2.239586486816406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,1,1,power_law_1.2,20.182835693359372
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,1,1,power_law_1.2,2.4527743530273436
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,1,1,power_law_1.2,2.694549865722656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,1,1,power_law_1.2,3.176295166015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,1,1,power_law_1.2,3.9013433837890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,1,1,power_law_1.2,4.4629364013671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,1,1,balanced,0.23335296630859376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,1,1,balanced,0.43402751922607424
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,1,1,power_law_1.2,6.124375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,1,1,power_law_1.2,26.08275146484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,1,1,balanced,0.7091027069091796
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,1,1,balanced,1.3397804260253907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,1,1,power_law_1.2,7.522655639648437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,1,1,balanced,2.7190643310546876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,1,1,balanced,5.220646362304687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,1,1,balanced,7.988969116210937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,1,1,power_law_1.2,11.076190185546874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,1,1,balanced,7.877219238281251
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,1,1,balanced,7.97651123046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,1,1,balanced,7.883306884765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,1,1,power_law_1.2,38.089206542968746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,1,1,balanced,7.941288452148437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,1,1,balanced,8.055695190429688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,1,1,balanced,8.003888549804689
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,1,1,balanced,8.096024780273437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,1,1,power_law_1.2,14.024058837890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,1,1,balanced,8.1540283203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,1,1,balanced,8.201444702148438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,1,1,balanced,12.68854736328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,1,1,balanced,13.02385498046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,1,1,balanced,13.458424072265625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,1,1,balanced,14.320986328124999
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,1,1,balanced,14.990321044921876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,1,1,power_law_1.2,50.4067236328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,1,1,balanced,0.12923839569091797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,1,1,balanced,0.22384511947631838
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,1,1,balanced,0.4047545623779297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,1,1,balanced,15.87456787109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,1,1,balanced,0.7683916473388672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,1,1,balanced,1.3942828369140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,1,1,balanced,2.5959481811523437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,1,1,balanced,17.911564941406247
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,1,1,balanced,3.8747161865234374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,1,1,balanced,3.890072326660156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,1,1,balanced,3.8800735473632812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,1,1,balanced,3.9476824951171876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,1,1,balanced,4.079324035644531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,1,1,balanced,4.049868774414063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,1,1,balanced,19.013162841796877
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,1,1,balanced,3.97841796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,1,1,balanced,4.031268615722657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,1,1,balanced,4.012288513183593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,1,1,balanced,4.043692016601563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,1,1,balanced,4.440205993652343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,1,1,balanced,4.517740783691407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,1,1,balanced,26.646608886718752
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,1,1,balanced,4.837452697753906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,1,1,balanced,5.092046203613281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,1,1,balanced,5.252805786132813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,1,1,balanced,5.518789672851563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,1,1,balanced,5.81915771484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,1,1,power_law_1.01,0.4505696105957031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,1,1,power_law_1.01,0.23207744598388672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,1,1,power_law_1.01,0.7063212585449219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,1,1,balanced,6.490120239257813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,1,1,power_law_1.01,1.3373536682128906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,1,1,power_law_1.01,2.4428121948242185
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,1,1,balanced,9.36976318359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,1,1,balanced,37.11944580078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,1,1,power_law_1.01,3.45340087890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,1,1,balanced,10.843052978515626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,1,1,power_law_1.01,3.978049926757812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,1,1,power_law_1.01,5.193208618164062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,1,1,balanced,14.211429443359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,1,1,power_law_1.01,5.4899359130859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,1,1,power_law_1.01,5.7574572753906255
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,1,1,power_law_1.01,6.0870776367187505
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,1,1,power_law_1.01,6.026656494140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,1,1,power_law_1.01,6.495766601562499
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,1,1,power_law_1.01,6.791803588867188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,1,1,power_law_1.01,7.408471069335937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,1,1,power_law_1.01,7.843367919921874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,1,1,power_law_1.01,10.641785888671874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,1,1,power_law_1.01,11.455021972656251
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,1,1,balanced,55.830473632812506
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,1,1,power_law_1.01,12.264036865234376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,1,1,power_law_1.01,13.540020751953126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,1,1,power_law_1.01,0.1288704013824463
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,1,1,power_law_1.01,14.408021240234376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,1,1,power_law_1.01,0.2236204719543457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,1,1,power_law_1.01,0.4041088104248047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,1,1,power_law_1.01,0.7665190124511719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,1,1,power_law_1.01,16.408428955078126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,1,1,power_law_1.01,1.2982073974609374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,1,1,power_law_1.01,1.7694873046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,1,1,power_law_1.01,2.031057891845703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,1,1,power_law_1.01,18.578863525390624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,1,1,power_law_1.01,2.5786297607421877
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,1,1,power_law_1.01,2.719824523925781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,1,1,power_law_1.01,2.8479086303710934
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,1,1,power_law_1.01,2.9489254760742187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,1,1,power_law_1.01,2.9753805541992184
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,1,1,power_law_1.01,3.097333679199219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,1,1,power_law_1.01,23.286550292968748
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,1,1,power_law_1.01,3.203720397949219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,1,1,power_law_1.01,3.2299737548828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,1,1,power_law_1.01,3.327271118164062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,1,1,power_law_1.01,3.6838452148437497
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,1,1,power_law_1.01,3.9105120849609376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,1,1,power_law_1.01,4.1199316406249995
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,1,1,power_law_1.01,4.4565478515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,1,1,power_law_1.01,28.14811279296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,1,1,power_law_1.01,4.750181884765626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,1,1,power_law_1.01,5.462689208984375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,1,1,power_law_1.2,0.23377664566040038
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,1,1,power_law_1.01,6.186746215820312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,1,1,power_law_1.2,0.4342073440551758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,1,1,power_law_1.2,0.7159085083007812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,1,1,power_law_1.01,7.468019409179687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,1,1,power_law_1.2,1.338170928955078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,1,1,power_law_1.01,9.051031494140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,1,1,power_law_1.2,2.2325689697265623
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,1,1,power_law_1.2,3.1523876953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,1,1,power_law_1.01,38.66212646484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,1,1,power_law_1.2,3.644234924316406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,1,1,power_law_1.01,11.973602294921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,1,1,power_law_1.2,5.084108276367187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,1,1,power_law_1.2,5.338056030273437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,1,1,power_law_1.01,14.694271240234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,1,1,power_law_1.2,5.6881738281250005
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,1,1,power_law_1.2,5.765351562499999
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,1,1,power_law_1.2,6.199052734375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,1,1,power_law_1.2,6.367576293945312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,1,1,power_law_1.01,48.75580078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,1,1,power_law_1.2,7.020264282226563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,1,1,power_law_1.2,7.398355712890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,1,1,power_law_1.2,7.92487548828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,1,1,power_law_1.2,10.80291015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,1,1,power_law_1.2,11.73241455078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,1,1,power_law_1.2,12.3892138671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,1,1,power_law_1.2,13.682506103515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,1,1,power_law_1.2,15.089936523437501
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,1,1,power_law_1.2,0.22378623962402341
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,1,1,power_law_1.2,0.40449153900146484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,1,1,power_law_1.2,17.207666015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,1,1,power_law_1.2,0.1287168025970459
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,1,1,power_law_1.2,0.7673363494873047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,1,1,power_law_1.2,1.2013478088378906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,1,1,power_law_1.2,1.6168333435058593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,1,1,power_law_1.2,19.56968017578125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,1,1,power_law_1.2,1.867073211669922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,1,1,power_law_1.2,2.530917053222656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,1,1,power_law_1.2,2.6401727294921877
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,1,1,power_law_1.2,2.7353170776367186
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,1,1,power_law_1.2,2.8419366455078126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,1,1,power_law_1.2,24.3331787109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,1,1,power_law_1.2,2.9598388671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,1,1,power_law_1.2,3.041777954101563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,1,1,power_law_1.2,3.140577392578125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,1,1,power_law_1.2,3.2223706054687495
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,1,1,power_law_1.2,3.2981491088867188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,1,1,power_law_1.2,3.763134765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,1,1,power_law_1.2,3.928895263671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,1,1,power_law_1.2,4.183205261230468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,1,1,power_law_1.2,4.712945861816406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,1,1,power_law_1.2,29.458515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,1,1,power_law_1.2,4.975416259765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,1,1,power_law_1.2,5.8754602050781255
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,1,1,balanced,0.08787391662597657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,1,1,balanced,0.13392191886901855
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,1,1,balanced,0.23594944000244142
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,1,1,power_law_1.2,6.529465942382812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,1,1,balanced,0.44198654174804686
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,1,1,balanced,0.8597203063964844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,1,1,balanced,1.6888243103027343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,1,1,power_law_1.2,7.772172241210937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,1,1,balanced,1.6888671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,1,1,balanced,1.7025549316406248
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,1,1,balanced,1.7275315856933595
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,1,1,balanced,1.7190585327148438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,1,1,balanced,1.730335388183594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,1,1,power_law_1.2,9.625712280273437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,1,1,power_law_1.2,39.723388671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,1,1,balanced,1.7495698547363283
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,1,1,balanced,1.8013011169433593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,1,1,balanced,1.780919647216797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,1,1,power_law_1.2,12.292233886718751
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,1,1,balanced,2.7225625610351565
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,1,1,balanced,2.682894592285156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,1,1,balanced,2.7864334106445314
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,1,1,balanced,2.943303527832031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,1,1,balanced,3.0646817016601564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,1,1,power_law_1.2,15.128331298828124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,1,1,balanced,3.2276242065429686
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,1,1,balanced,3.338597717285156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,1,1,balanced,0.05975232124328613
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,1,1,balanced,3.8203079223632814
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,1,1,balanced,0.09442111968994141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,1,1,balanced,4.0932620239257815
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,1,1,balanced,0.14362303733825682
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,1,1,balanced,0.2543257522583008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,1,1,balanced,0.44856895446777345
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,1,1,balanced,0.8629535675048828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,1,1,balanced,0.8612319946289062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,1,1,balanced,0.8883712005615234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,1,1,balanced,0.8876672363281249
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,1,1,balanced,5.8973455810546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,1,1,balanced,0.8765510559082031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,1,1,balanced,0.9064825439453126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,1,1,balanced,0.9173056030273438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,1,1,balanced,0.9284403228759766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,1,1,balanced,0.9093497467041015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,1,1,balanced,1.0386988830566406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,1,1,balanced,7.5949816894531255
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,1,1,power_law_1.2,49.89626953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,1,1,balanced,1.073523178100586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,1,1,balanced,1.10740478515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,1,1,balanced,1.0633363342285156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,1,1,balanced,1.092194595336914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,1,1,balanced,1.2005337524414064
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,1,1,balanced,1.2360441589355469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,1,1,balanced,1.453324737548828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,1,1,power_law_1.01,0.0833779239654541
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,1,1,balanced,1.336505584716797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,1,1,balanced,2.198916473388672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,1,1,power_law_1.01,0.13174400329589844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,1,1,power_law_1.01,0.2363814353942871
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,1,1,balanced,2.4274981689453123
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,1,1,power_law_1.01,0.4544780731201172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,1,1,power_law_1.01,0.6690284729003906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,1,1,balanced,11.034182128906249
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,1,1,power_law_1.01,0.863333740234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,1,1,balanced,3.5080499267578125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,1,1,power_law_1.01,1.2278105926513672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,1,1,power_law_1.01,1.2959910583496093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,1,1,power_law_1.01,1.3279533386230469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,1,1,power_law_1.01,1.3426541137695314
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,1,1,power_law_1.01,1.4240039062499998
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,1,1,balanced,4.562630310058593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,1,1,power_law_1.01,1.4844947814941407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,1,1,power_law_1.01,1.5226541137695313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,1,1,power_law_1.01,1.5930674743652342
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,1,1,balanced,14.662335205078126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,1,1,power_law_1.01,2.3614604187011716
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,1,1,power_law_1.01,2.4012281799316404
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,1,1,power_law_1.01,2.530509490966797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,1,1,power_law_1.01,2.714871826171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,1,1,power_law_1.01,2.8798944091796876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,1,1,power_law_1.01,3.1261370849609373
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,1,1,power_law_1.01,3.351734924316406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,1,1,power_law_1.01,0.060195198059082036
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,1,1,power_law_1.01,0.09675200462341309
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,1,1,power_law_1.01,0.1530515193939209
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,1,1,power_law_1.01,0.2543168067932129
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,1,1,power_law_1.01,4.0721438598632815
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,1,1,power_law_1.01,0.35767742156982424
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,1,1,power_law_1.01,0.44966209411621094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,1,1,power_law_1.01,4.890071105957031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,1,1,power_law_1.01,0.6418380737304688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,1,1,power_law_1.01,0.6693452453613281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,1,1,power_law_1.01,0.6857350158691407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,1,1,power_law_1.01,0.733034896850586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,1,1,power_law_1.01,0.696929931640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,1,1,power_law_1.01,0.7556396484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,1,1,power_law_1.01,6.3381542968749995
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,1,1,power_law_1.01,0.7652210998535156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,1,1,power_law_1.01,0.7819660949707031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,1,1,power_law_1.01,0.8878105926513673
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,1,1,power_law_1.01,0.8928755187988282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,1,1,power_law_1.01,0.938936996459961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,1,1,power_law_1.01,0.9692940521240235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,1,1,power_law_1.01,1.052373733520508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,1,1,power_law_1.01,8.13961669921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,1,1,power_law_1.01,1.1734425354003908
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,1,1,power_law_1.01,1.260855026245117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,1,1,power_law_1.01,1.5745401000976562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,1,1,power_law_1.2,0.08016063690185546
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,1,1,power_law_1.01,1.7903724670410157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,1,1,power_law_1.2,0.13156224250793458
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,1,1,power_law_1.2,0.2400588798522949
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,1,1,power_law_1.01,2.3631053161621094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,1,1,power_law_1.2,0.4175372695922851
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,1,1,power_law_1.01,2.906218872070313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,1,1,power_law_1.2,0.6177740859985351
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,1,1,power_law_1.01,11.354205322265624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,1,1,power_law_1.2,0.8126445007324218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,1,1,power_law_1.01,3.994689331054688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,1,1,power_law_1.2,1.2021920013427736
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,1,1,power_law_1.2,1.2765408325195313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,1,1,power_law_1.2,1.302981719970703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,1,1,power_law_1.2,1.349614715576172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,1,1,power_law_1.01,5.169530029296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,1,1,power_law_1.2,1.4485498046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,1,1,power_law_1.2,1.480328369140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,1,1,power_law_1.2,1.5571160888671876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,1,1,power_law_1.2,1.6319168090820313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,1,1,power_law_1.01,14.376625976562499
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,1,1,power_law_1.2,2.405996856689453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,1,1,power_law_1.2,2.474343109130859
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,1,1,power_law_1.2,2.569045104980469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,1,1,power_law_1.2,2.777751770019531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,1,1,power_law_1.2,2.942009582519531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,1,1,power_law_1.2,3.2255975341796876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,1,1,power_law_1.2,3.4337850952148434
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,1,1,power_law_1.2,0.05677504062652587
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,1,1,power_law_1.2,0.09505279541015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,1,1,power_law_1.2,0.14141247749328614
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,1,1,power_law_1.2,0.2390950393676758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,1,1,power_law_1.2,0.32487873077392576
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,1,1,power_law_1.2,4.256718139648437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,1,1,power_law_1.2,0.42212863922119137
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,1,1,power_law_1.2,0.6230796813964844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,1,1,power_law_1.2,0.6574400329589845
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,1,1,power_law_1.2,5.0059640502929685
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,1,1,power_law_1.2,0.6731021118164062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,1,1,power_law_1.2,0.699912338256836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,1,1,power_law_1.2,0.7221868896484376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,1,1,power_law_1.2,0.7476179504394531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,1,1,power_law_1.2,0.764938201904297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,1,1,power_law_1.2,0.7941248321533203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,1,1,power_law_1.2,0.9079782104492187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,1,1,power_law_1.2,6.605914916992188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,1,1,power_law_1.2,0.9334521484374999
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,1,1,power_law_1.2,0.9309772491455078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,1,1,power_law_1.2,1.0167411041259764
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,1,1,power_law_1.2,1.0948665618896485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,1,1,power_law_1.2,1.24442626953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,1,1,power_law_1.2,1.4070687866210938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,1,1,power_law_1.2,8.316619262695312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,1,1,power_law_1.2,1.5668505859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,1,1,power_law_1.2,1.9120512390136717
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,1,1,power_law_1.2,2.3973965454101562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,1,1,power_law_1.2,2.9206814575195312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,1,1,power_law_1.2,11.454932861328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,1,1,power_law_1.2,3.965015563964844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,1,1,power_law_1.2,5.110807800292969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,1,1,power_law_1.2,14.687381591796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,1,1,balanced,0.07001471996307373
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,1,1,balanced,0.9232895660400391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,1,1,balanced,0.935203857421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,1,1,balanced,0.10892479896545411
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,1,1,balanced,0.21226623535156247
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,1,1,balanced,0.9249983978271483
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,1,1,balanced,0.9438995361328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,1,1,balanced,0.3704217529296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,1,1,balanced,0.6728800201416016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,1,1,balanced,0.9271456146240235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,1,1,balanced,0.9177536010742188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,1,1,balanced,1.3928230285644532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,1,1,balanced,1.4201242065429687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,1,1,balanced,1.4482643127441406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,1,1,balanced,1.4695666503906248
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,1,1,balanced,1.5121714782714843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,1,1,balanced,1.8788006591796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,1,1,balanced,1.5548141479492188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,1,1,balanced,2.0054393005371094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,1,1,balanced,1.6567327880859373
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,1,1,balanced,1.710853729248047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,1,1,balanced,2.774320068359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,1,1,power_law_1.01,0.06975296020507812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,1,1,balanced,3.102153015136719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,1,1,balanced,4.646127319335937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,1,1,power_law_1.01,0.10885055541992188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,1,1,power_law_1.01,0.7371430206298828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,1,1,balanced,5.885099487304688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,1,1,power_law_1.01,0.7803020477294922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,1,1,power_law_1.01,0.8285228729248046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,1,1,power_law_1.01,0.6974931335449218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,1,1,power_law_1.01,0.8523865509033204
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,1,1,balanced,8.7643798828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,1,1,power_law_1.01,0.8125856018066406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,1,1,power_law_1.01,1.2908773803710938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,1,1,power_law_1.01,1.3324237060546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,1,1,power_law_1.01,1.380585021972656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,1,1,power_law_1.01,0.33947841644287113
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,1,1,power_law_1.01,0.21213312149047853
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,1,1,power_law_1.01,1.4156358337402344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,1,1,power_law_1.01,1.4447917175292968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,1,1,power_law_1.01,0.44763648986816407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,1,1,power_law_1.01,1.5146925354003906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,1,1,power_law_1.01,1.6095993041992187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,1,1,balanced,11.541878662109374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,1,1,power_law_1.01,1.703570556640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,1,1,power_law_1.01,1.9278060913085937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,1,1,power_law_1.01,2.18383544921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,1,1,power_law_1.01,2.7375604248046876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,1,1,power_law_1.01,3.368331604003906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,1,1,power_law_1.01,4.626971435546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,1,1,power_law_1.2,0.10707072257995605
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,1,1,power_law_1.2,0.21100160598754886
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,1,1,power_law_1.2,0.40784320831298826
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,1,1,power_law_1.2,0.6932332611083984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,1,1,power_law_1.01,6.109550170898437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,1,1,power_law_1.2,0.7470502471923828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,1,1,power_law_1.2,0.2946316719055176
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,1,1,power_law_1.2,0.06961152076721192
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,1,1,power_law_1.2,0.8225548553466797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,1,1,power_law_1.2,0.8675078582763671
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,1,1,power_law_1.2,1.3335244750976563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,1,1,power_law_1.2,1.3630029296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,1,1,power_law_1.2,1.4101434326171876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,1,1,power_law_1.2,1.4554432678222657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,1,1,power_law_1.01,8.5837939453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,1,1,power_law_1.2,0.7924614715576171
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,1,1,power_law_1.2,1.4956997680664064
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,1,1,power_law_1.2,1.5410585021972656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,1,1,power_law_1.2,0.8424313354492188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,1,1,power_law_1.2,1.63553466796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,1,1,power_law_1.2,1.700256652832031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,1,1,power_law_1.2,1.9749229431152344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,1,1,power_law_1.2,2.251604461669922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,1,1,power_law_1.01,11.333197021484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,1,1,power_law_1.2,2.833900146484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,1,1,power_law_1.2,3.4573785400390626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,1,1,power_law_1.2,4.670627746582031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,1,1,balanced,0.7260160064697265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,1,1,balanced,1.3805107116699218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,1,1,power_law_1.2,5.926224975585937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,1,1,balanced,2.692310485839844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,1,1,balanced,0.3939628982543945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,1,1,power_law_1.2,8.514149169921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,1,1,balanced,5.2712335205078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,1,1,balanced,7.763148803710938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,1,1,balanced,7.760806884765626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,1,1,power_law_1.2,11.327266845703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,1,1,balanced,7.81884033203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,1,1,balanced,7.857340698242187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,1,1,balanced,7.913791503906251
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,1,1,balanced,7.940430908203124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,1,1,balanced,8.031716918945312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,1,1,balanced,8.092020263671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,1,1,balanced,8.158137817382812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,1,1,balanced,8.219247436523437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,1,1,balanced,10.908101806640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,1,1,balanced,13.97003173828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,1,1,balanced,14.32701171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,1,1,balanced,14.892027587890624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,1,1,balanced,16.705369873046873
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,1,1,balanced,17.412906494140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,1,1,balanced,0.42652286529541017
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,1,1,balanced,0.7933267211914062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,1,1,balanced,25.26460693359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,1,1,balanced,0.23785280227661132
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,1,1,balanced,29.229062499999998
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,1,1,balanced,1.4246617126464844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,1,1,balanced,2.6376312255859373
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,1,1,balanced,3.8702624511718753
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,1,1,balanced,4.243507080078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,1,1,balanced,3.88321533203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,1,1,balanced,3.9171942138671874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,1,1,balanced,3.9361325073242184
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,1,1,balanced,4.003811340332031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,1,1,balanced,4.026095275878906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,1,1,balanced,4.170723876953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,1,1,balanced,4.255368347167969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,1,1,balanced,47.0300341796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,1,1,balanced,4.2130258178710935
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,1,1,balanced,4.1906124877929685
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,1,1,balanced,4.308612365722657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,1,1,balanced,4.7374987792968755
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,1,1,balanced,4.889049682617188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,1,1,balanced,5.21506103515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,1,1,balanced,59.1785107421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,1,1,balanced,5.454979248046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,1,1,balanced,9.114871215820312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,1,1,balanced,9.370348510742186
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,1,1,power_law_1.01,0.730972137451172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,1,1,power_law_1.01,0.41695232391357423
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,1,1,power_law_1.01,1.3667301940917969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,1,1,balanced,13.535440673828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,1,1,balanced,14.917320556640624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,1,1,power_law_1.01,2.4228883361816407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,1,1,power_law_1.01,3.433971252441406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,1,1,power_law_1.01,5.7290716552734375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,1,1,power_law_1.01,6.187032470703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,1,1,power_law_1.01,6.823724975585938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,1,1,power_law_1.01,6.963921508789062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,1,1,power_law_1.01,6.858876953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,1,1,balanced,27.354501953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,1,1,power_law_1.01,7.2286267089843745
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,1,1,balanced,90.04923828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,1,1,balanced,22.697866210937498
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,1,1,power_law_1.01,7.578899536132812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,1,1,power_law_1.01,8.180708618164063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,1,1,power_law_1.01,9.536536254882813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,1,1,power_law_1.01,10.37642822265625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,1,1,power_law_1.01,11.66322021484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,1,1,power_law_1.01,13.538834228515626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,1,1,power_law_1.01,14.06096923828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,1,1,power_law_1.01,14.88302001953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,1,1,power_law_1.01,16.56198486328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,1,1,power_law_1.01,18.77539306640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,1,1,balanced,112.942177734375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,1,1,power_law_1.01,0.422242546081543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,1,1,power_law_1.01,23.67551513671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,1,1,power_law_1.01,0.7919609832763672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,1,1,power_law_1.01,0.23568639755249024
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,1,1,power_law_1.01,28.9268603515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,1,1,power_law_1.01,1.3029702758789061
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,1,1,power_law_1.01,1.767957763671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,1,1,power_law_1.01,2.8657183837890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,1,1,power_law_1.01,3.2249575805664064
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,1,1,power_law_1.01,3.3710797119140623
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,1,1,power_law_1.01,3.4674206542968746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,1,1,power_law_1.01,3.0736703491210937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,1,1,power_law_1.01,38.84667236328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,1,1,power_law_1.01,3.3190292358398437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,1,1,power_law_1.01,3.553282470703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,1,1,power_law_1.01,3.6219122314453123
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,1,1,power_law_1.01,3.742695617675781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,1,1,power_law_1.01,3.7784875488281253
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,1,1,power_law_1.01,3.911479187011719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,1,1,power_law_1.01,4.148560791015624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,1,1,power_law_1.01,4.74935302734375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,1,1,power_law_1.01,5.892049560546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,1,1,power_law_1.01,50.23931640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,1,1,power_law_1.01,7.947429809570312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,1,1,power_law_1.01,4.923606262207032
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,1,1,power_law_1.01,9.79361083984375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,1,1,power_law_1.01,6.31205078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,1,1,power_law_1.2,0.3981907272338867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,1,1,power_law_1.2,1.3670034790039063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,1,1,power_law_1.01,12.665050048828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,1,1,power_law_1.2,0.7243328094482422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,1,1,power_law_1.2,2.226908874511719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,1,1,power_law_1.01,15.80989501953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,1,1,power_law_1.2,3.0512115478515627
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,1,1,power_law_1.2,5.599009399414062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,1,1,power_law_1.2,6.305111083984375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,1,1,power_law_1.2,6.4415771484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,1,1,power_law_1.01,73.862421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,1,1,power_law_1.01,22.0639990234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,1,1,power_law_1.2,6.886510009765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,1,1,power_law_1.2,6.991471557617187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,1,1,power_law_1.2,7.466609497070313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,1,1,power_law_1.2,8.036826782226562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,1,1,power_law_1.2,8.724312133789063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,1,1,power_law_1.01,28.38348388671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,1,1,power_law_1.2,10.106877441406251
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,1,1,power_law_1.2,11.000196533203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,1,1,power_law_1.2,11.724180908203126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,1,1,power_law_1.2,14.246499023437499
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,1,1,power_law_1.2,14.505494384765626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,1,1,power_law_1.2,15.326959228515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,1,1,power_law_1.01,97.01877929687501
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,1,1,power_law_1.2,17.191429443359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,1,1,power_law_1.2,19.5602001953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,1,1,power_law_1.2,0.42354816436767584
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,1,1,power_law_1.2,0.23578943252563475
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,1,1,power_law_1.2,0.7925279998779298
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,1,1,power_law_1.2,24.415422363281248
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,1,1,power_law_1.2,2.8246841430664062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,1,1,power_law_1.2,1.199046401977539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,1,1,power_law_1.2,1.5781170654296877
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,1,1,power_law_1.2,29.5277294921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,1,1,power_law_1.2,3.060754699707031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,1,1,power_law_1.2,3.18278076171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,1,1,power_law_1.2,3.3072039794921877
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,1,1,power_law_1.2,3.4168267822265626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,1,1,power_law_1.2,3.5197918701171873
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,1,1,power_law_1.2,3.6370648193359374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,1,1,power_law_1.2,3.7387585449218745
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,1,1,power_law_1.2,3.87225341796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,1,1,power_law_1.2,4.012666931152344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,1,1,power_law_1.2,4.1355712890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,1,1,power_law_1.2,4.309722290039063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,1,1,power_law_1.2,5.064641418457031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,1,1,power_law_1.2,5.246362915039063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,1,1,power_law_1.2,41.2681591796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,1,1,power_law_1.2,6.112115478515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,1,1,power_law_1.2,6.731538085937499
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,1,1,power_law_1.2,8.133075561523437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,1,1,power_law_1.2,9.710565795898438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,1,2,balanced,0.18672128677368166
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,1,1,power_law_1.2,13.263527832031249
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,1,2,balanced,0.27686784744262694
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,1,2,balanced,0.5713062286376953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,1,2,balanced,0.5730879974365235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,1,2,balanced,0.5781619262695312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,1,1,power_law_1.2,53.4243310546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,1,2,balanced,0.6531545257568359
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,1,2,balanced,0.5959987258911132
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,1,2,balanced,0.43773056030273433
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,1,2,balanced,0.6050764846801757
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,1,2,balanced,0.6131814575195312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,1,2,balanced,0.6206604766845703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,1,2,balanced,0.6402841949462891
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,1,2,balanced,0.6460684967041016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,1,1,power_law_1.2,16.309505615234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,1,2,balanced,0.8027776336669922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,1,2,balanced,0.7650777435302734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,1,2,balanced,0.6643212890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,1,2,balanced,1.2262105560302734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,1,2,balanced,1.50115966796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,1,2,balanced,0.8370111846923829
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,1,2,balanced,2.2400857543945314
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,1,2,balanced,2.787145080566406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,1,2,balanced,4.246305236816406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,1,2,balanced,5.648245849609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,1,2,balanced,8.936409301757813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,1,2,balanced,0.10465663909912108
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,1,2,balanced,0.13744511604309081
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,1,2,balanced,12.511411132812501
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,1,1,power_law_1.2,22.15079345703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,1,2,balanced,0.23159551620483398
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,1,2,balanced,0.23193344116210937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,1,2,balanced,0.2325619125366211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,1,2,balanced,0.2341324806213379
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,1,2,balanced,0.2359231948852539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,1,2,balanced,0.24636287689208985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,1,2,balanced,0.23794559478759764
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,1,2,balanced,0.2456332778930664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,1,2,balanced,0.24057600021362308
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,1,2,balanced,0.24937728881835936
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,1,2,balanced,0.25690879821777346
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,1,2,balanced,0.26695039749145505
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,1,2,balanced,0.2722559928894043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,1,2,balanced,0.20834688186645506
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,1,2,balanced,0.27657215118408207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,1,2,balanced,0.3598643112182617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,1,2,balanced,0.3909900665283203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,1,2,balanced,0.5991104125976563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,1,2,balanced,0.7332147216796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,1,2,balanced,1.0461734771728515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,1,2,balanced,1.5183091735839844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,1,2,balanced,2.2819059753417967
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,1,2,balanced,2.9022872924804686
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,1,2,power_law_1.01,0.24595584869384765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,1,2,balanced,19.99526611328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,1,2,power_law_1.01,0.26765695571899417
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,1,2,power_law_1.01,0.2918592071533203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,1,2,balanced,4.520834655761719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,1,2,balanced,5.756483764648438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,1,2,power_law_1.01,0.5749734497070312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,1,2,power_law_1.01,0.575810546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,1,2,power_law_1.01,0.6054502487182617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,1,2,power_law_1.01,0.6224307250976563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,1,2,power_law_1.01,0.6366451263427735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,1,2,power_law_1.01,0.5879359817504882
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,1,2,power_law_1.01,0.6648934173583985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,1,2,power_law_1.01,0.6625676727294921
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,1,2,power_law_1.01,0.7065510559082031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,1,2,power_law_1.01,0.8057421112060548
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,1,2,power_law_1.01,0.9097164916992189
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,1,2,power_law_1.01,0.9577779388427734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,1,2,power_law_1.01,0.5988224029541016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,1,2,balanced,27.038693847656248
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,1,2,power_law_1.01,1.2105228424072265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,1,1,power_law_1.2,78.327333984375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,1,2,power_law_1.01,1.6689599609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,1,2,power_law_1.01,2.2841445922851564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,1,2,power_law_1.01,3.040097351074219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,1,2,power_law_1.01,6.231064453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,1,1,power_law_1.2,28.19214599609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,1,2,power_law_1.01,3.982077331542969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,1,2,power_law_1.01,7.573857421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,1,2,power_law_1.01,0.1316697597503662
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,1,2,power_law_1.01,0.11649663925170899
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,1,2,power_law_1.01,13.572103271484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,1,2,power_law_1.01,0.1479360008239746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,1,2,power_law_1.01,18.005133056640624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,1,2,power_law_1.01,0.22660352706909181
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,1,2,power_law_1.01,0.23577600479125976
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,1,2,power_law_1.01,0.22809343338012694
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,1,2,power_law_1.01,0.23781759262084962
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,1,2,power_law_1.01,0.24236799240112306
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,1,2,power_law_1.01,0.25686016082763674
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,1,2,power_law_1.01,0.26181631088256835
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,1,2,power_law_1.01,0.23002239227294924
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,1,2,power_law_1.01,0.27597951889038086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,1,2,power_law_1.01,0.29410432815551757
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,1,2,power_law_1.01,0.3008537673950195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,1,2,power_law_1.01,0.34190078735351564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,1,2,power_law_1.01,0.4996697616577149
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,1,2,power_law_1.01,0.5666726303100587
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,1,2,power_law_1.01,0.2320307159423828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,1,2,power_law_1.01,0.8418048095703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,1,2,power_law_1.01,1.0619840240478515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,1,2,power_law_1.01,2.0230438232421877
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,1,2,power_law_1.01,1.5468710327148438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,1,2,power_law_1.01,0.23387392044067382
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,1,2,power_law_1.01,3.143240966796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,1,2,power_law_1.2,0.3142336082458496
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,1,2,power_law_1.01,4.018557434082031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,1,2,power_law_1.2,0.2340185546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,1,2,power_law_1.2,0.2709542465209961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,1,2,power_law_1.2,0.5819635009765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,1,1,power_law_1.2,101.74196289062499
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,1,2,power_law_1.2,0.5743193435668945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,1,2,power_law_1.2,0.594598388671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,1,2,power_law_1.2,0.6287295913696289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,1,2,power_law_1.2,0.646398696899414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,1,2,power_law_1.2,0.5764428710937499
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,1,2,power_law_1.2,0.662192611694336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,1,2,power_law_1.2,0.6774681854248047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,1,2,power_law_1.2,0.7050265502929688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,1,2,power_law_1.2,0.7927513885498046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,1,2,power_law_1.2,0.9303731536865234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,1,2,power_law_1.2,0.960211181640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,1,2,power_law_1.01,22.595747070312502
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,1,2,power_law_1.2,1.235077133178711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,1,2,power_law_1.2,1.6119847106933594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,1,2,power_law_1.01,5.395062866210937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,1,2,power_law_1.2,0.6804940795898438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,1,2,power_law_1.2,2.296913909912109
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,1,2,power_law_1.2,3.0093286132812502
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,1,2,power_law_1.2,3.884408264160156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,1,2,power_law_1.01,35.707431640624996
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,1,2,power_law_1.2,0.10802559852600098
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,1,2,power_law_1.2,0.1469491195678711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,1,2,power_law_1.2,6.287739868164063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,1,2,power_law_1.2,0.22780672073364258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,1,2,power_law_1.2,0.22433536529541015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,1,2,power_law_1.2,0.2311961555480957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,1,2,power_law_1.2,0.2328358459472656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,1,2,power_law_1.2,0.234835205078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,1,2,power_law_1.2,7.494472656249999
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,1,2,power_law_1.2,0.23645824432373047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,1,2,power_law_1.2,0.24219520568847655
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,1,2,power_law_1.2,0.24267391204833982
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,1,2,power_law_1.2,0.2603840065002442
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,1,2,power_law_1.2,0.26222208023071286
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,1,2,power_law_1.2,0.2743462371826172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,1,2,power_law_1.2,0.2927103996276855
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,1,2,power_law_1.2,0.3052582359313965
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,1,2,power_law_1.2,0.3417446517944336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,1,2,power_law_1.2,0.49036926269531256
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,1,2,power_law_1.2,0.5748582458496093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,1,2,power_law_1.2,0.1627289581298828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,1,2,power_law_1.2,0.8361881256103516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,1,2,power_law_1.2,1.1556082916259764
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,1,2,power_law_1.2,13.842442626953124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,1,2,power_law_1.01,7.972442626953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,1,2,power_law_1.2,1.8382284545898435
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,1,2,power_law_1.2,18.4683349609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,1,2,power_law_1.2,3.0564248657226565
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,1,2,power_law_1.2,1.4641651916503906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,1,2,power_law_1.2,4.127196044921876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,2,1,balanced,0.35530174255371094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,2,1,balanced,0.11410495758056641
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,2,1,balanced,0.3561305618286133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,2,1,balanced,0.20155263900756837
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,1,2,power_law_1.2,5.662003173828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,2,1,balanced,0.3564352035522461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,2,1,balanced,0.3550796890258789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,2,1,balanced,0.3678771209716797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,2,1,balanced,0.3703583908081055
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,2,1,balanced,0.36320575714111325
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,2,1,balanced,0.37358463287353516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,2,1,balanced,0.4123091125488282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,2,1,balanced,0.4178028869628906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,2,1,balanced,0.4268371200561523
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,2,1,balanced,0.43870590209960936
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,2,1,balanced,0.36787071228027346
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,2,1,balanced,0.3685094451904297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,2,1,balanced,0.9671981048583984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,2,1,balanced,0.36469310760498047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,2,1,balanced,1.2446444702148436
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,1,2,power_law_1.2,22.41723876953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,2,1,balanced,0.6237356948852539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,2,1,balanced,0.665315170288086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,1,2,power_law_1.2,8.43236328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,2,1,balanced,1.9179257202148439
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,2,1,balanced,2.464427490234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,2,1,balanced,3.6973184204101566
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,2,1,balanced,0.0831827163696289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,2,1,balanced,0.20315776824951173
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,2,1,balanced,4.832008361816406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,1,2,power_law_1.2,37.97197021484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,2,1,balanced,0.204956169128418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,2,1,balanced,0.20247871398925782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,2,1,balanced,0.12092479705810547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,2,1,balanced,7.354100952148437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,2,1,balanced,0.20505344390869143
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,2,1,balanced,0.20673088073730467
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,2,1,balanced,0.22650495529174805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,2,1,balanced,0.2840300750732422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,2,1,balanced,0.28916479110717774
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,2,1,balanced,0.2080620765686035
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,2,1,balanced,10.175511474609376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,2,1,balanced,0.2096089553833008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,2,1,balanced,0.2569132804870605
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,2,1,balanced,0.34597633361816404
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,2,1,balanced,0.35830078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,2,1,balanced,0.21057855606079104
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,2,1,balanced,0.21411008834838868
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,2,1,balanced,1.3268511962890623
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,2,1,balanced,0.21733119964599606
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,2,1,balanced,2.0686579895019532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,2,1,balanced,0.22031744003295897
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,2,1,balanced,2.6545611572265626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,2,1,balanced,4.183137207031249
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,2,1,balanced,5.414186401367187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,2,1,power_law_1.01,0.22907072067260742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,2,1,power_law_1.01,0.11353471755981445
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,2,1,balanced,0.5320191955566406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,2,1,power_law_1.01,0.37532543182373046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,2,1,balanced,0.6900025939941405
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,2,1,balanced,1.0481804656982423
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,2,1,power_law_1.01,0.37703041076660154
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,2,1,power_law_1.01,0.44110656738281245
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,2,1,power_law_1.01,0.48171646118164063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,2,1,power_law_1.01,0.4938700866699219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,2,1,power_law_1.01,0.19368127822875975
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,2,1,power_law_1.01,0.3475596618652344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,2,1,power_law_1.01,0.5082175827026367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,2,1,power_law_1.01,0.5299961471557617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,2,1,power_law_1.01,0.3544313430786133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,2,1,power_law_1.01,0.7528460693359376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,2,1,power_law_1.01,0.42172607421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,2,1,power_law_1.01,0.432537612915039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,2,1,power_law_1.01,0.4165516662597656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,2,1,power_law_1.01,1.171189727783203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,2,1,power_law_1.01,1.5040447998046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,2,1,power_law_1.01,0.3611616134643555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,2,1,power_law_1.01,0.5270105743408203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,2,1,power_law_1.01,0.8854124450683594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,2,1,power_law_1.01,3.886995849609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,2,1,power_law_1.01,2.097775421142578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,2,1,power_law_1.01,0.2044723129272461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,2,1,power_law_1.01,2.660993957519531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,2,1,power_law_1.01,0.20744064331054685
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,2,1,power_law_1.01,0.07838143825531005
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,2,1,power_law_1.01,0.1380070400238037
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,2,1,power_law_1.01,0.20406719207763674
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,2,1,power_law_1.01,0.21410816192626952
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,2,1,power_law_1.01,0.2311065673828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,2,1,power_law_1.01,0.2025593566894531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,2,1,power_law_1.01,0.239880313873291
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,2,1,power_law_1.01,0.25396032333374025
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,2,1,power_law_1.01,0.23166656494140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,2,1,power_law_1.01,0.11722111701965332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,2,1,power_law_1.01,0.27607744216918945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,2,1,power_law_1.01,0.2998111915588379
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,2,1,power_law_1.01,0.3937900924682617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,2,1,power_law_1.01,0.2066348838806152
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,2,1,power_law_1.01,0.440081901550293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,2,1,power_law_1.01,7.4914501953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,2,1,power_law_1.01,5.0398797607421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,2,1,power_law_1.01,0.2049491119384766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,2,1,power_law_1.01,1.1087833404541017
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,2,1,power_law_1.01,0.28538688659667966
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,2,1,power_law_1.01,2.1832147216796876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,2,1,power_law_1.01,0.6124243164062501
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,2,1,power_law_1.01,0.7408723449707031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,2,1,power_law_1.2,0.20866880416870118
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,2,1,power_law_1.01,2.8485882568359377
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,2,1,power_law_1.01,1.4722271728515626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,2,1,power_law_1.2,0.11272383689880372
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,2,1,power_law_1.2,0.33874046325683593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,2,1,power_law_1.2,0.3483353424072265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,2,1,power_law_1.2,0.4171193695068359
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,2,1,power_law_1.2,0.4402323150634766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,2,1,power_law_1.2,0.3734790420532227
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,2,1,power_law_1.2,0.42834304809570317
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,2,1,power_law_1.2,0.18552383422851562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,2,1,power_law_1.2,0.4401011276245117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,2,1,power_law_1.01,10.340069580078126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,2,1,power_law_1.2,0.48272960662841796
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,2,1,power_law_1.2,0.3984812927246094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,2,1,power_law_1.01,4.267086181640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,2,1,power_law_1.2,0.5117971038818359
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,2,1,power_law_1.2,0.5503647994995118
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,2,1,power_law_1.2,0.5289228820800782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,2,1,power_law_1.2,0.3619180679321289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,2,1,power_law_1.2,0.6926656341552734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,2,1,power_law_1.2,0.848160629272461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,2,1,power_law_1.2,1.1156307220458985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,2,1,power_law_1.2,1.4488236999511719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,2,1,power_law_1.01,5.524950561523438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,2,1,power_law_1.2,0.08098112106323242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,2,1,power_law_1.2,0.48996864318847655
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,2,1,power_law_1.2,2.6601959228515626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,2,1,power_law_1.2,2.088106231689453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,2,1,power_law_1.2,0.19882368087768554
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,2,1,power_law_1.2,3.9074072265625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,2,1,power_law_1.2,0.20577728271484377
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,2,1,power_law_1.2,0.20110719680786132
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,2,1,power_law_1.2,0.20949440002441405
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,2,1,power_law_1.2,0.11524864196777344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,2,1,power_law_1.2,0.12433343887329103
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,2,1,power_law_1.2,0.21417152404785159
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,2,1,power_law_1.2,0.23210687637329103
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,2,1,power_law_1.2,0.23645759582519532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,2,1,power_law_1.2,0.23649152755737304
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,2,1,power_law_1.2,0.2527801513671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,2,1,power_law_1.2,0.2715315246582031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,2,1,power_law_1.2,0.2909644889831543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,2,1,power_law_1.2,0.3769894409179687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,2,1,power_law_1.2,0.4377497482299805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,2,1,power_law_1.2,0.20663871765136718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,2,1,power_law_1.2,0.20310144424438475
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,2,1,power_law_1.2,5.087848205566407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,2,1,power_law_1.2,0.5898777770996093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,2,1,power_law_1.2,1.1674700927734374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,2,1,power_law_1.2,0.29717311859130857
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,2,1,power_law_1.2,1.4349005126953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,1,2,balanced,0.24653440475463867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,1,2,balanced,0.3989606475830078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,1,2,balanced,0.6167027282714843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,2,1,power_law_1.2,0.7334508514404297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,1,2,balanced,0.9311603546142578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,1,2,balanced,0.9426367950439453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,2,1,power_law_1.2,2.184918975830078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,1,2,balanced,0.9444992065429687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,1,2,balanced,0.9615936279296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,1,2,balanced,0.9674662780761718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,1,2,balanced,0.9831449890136719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,2,1,power_law_1.2,7.425627441406251
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,1,2,balanced,0.9852428436279297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,1,2,balanced,1.0073228454589844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,1,2,balanced,1.0273267364501952
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,1,2,balanced,1.06495361328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,1,2,balanced,1.045765151977539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,1,2,balanced,1.454833984375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,1,2,balanced,1.3331674194335936
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,1,2,balanced,1.4425958251953124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,1,2,balanced,2.196697540283203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,1,2,balanced,2.594132385253906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,2,1,power_law_1.2,2.8529458618164063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,1,2,balanced,3.838535766601562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,1,2,balanced,4.681912231445312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,1,2,balanced,7.3223461914062495
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,1,2,balanced,10.193796997070312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,1,2,balanced,0.1218393611907959
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,1,2,balanced,0.1901811218261719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,1,2,balanced,0.32410240173339844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,1,2,balanced,0.33787902832031247
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,1,2,balanced,0.3386841583251953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,1,2,balanced,0.34117374420166013
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,2,1,power_law_1.2,10.275423583984374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,1,2,balanced,0.34301185607910156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,1,2,balanced,0.35715198516845703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,1,2,balanced,0.34926719665527345
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,1,2,balanced,0.3508697509765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,1,2,balanced,0.3539788818359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,1,2,balanced,0.3592319869995117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,1,2,balanced,0.3760460662841797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,1,2,balanced,0.38759552001953124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,1,2,balanced,0.3996390533447266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,1,2,balanced,0.366901741027832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,1,2,balanced,15.549163818359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,1,2,balanced,0.4079193496704102
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,1,2,balanced,0.5844147109985351
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,1,2,balanced,0.6350233459472656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,1,2,balanced,0.9584703826904297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,1,2,balanced,1.1645606231689452
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,1,2,balanced,1.8437338256835936
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,1,2,balanced,2.349827880859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,2,1,power_law_1.2,4.273336181640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,1,2,balanced,3.529477233886719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,1,2,power_law_1.01,0.28321151733398436
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,1,2,power_law_1.01,0.4633919906616211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,1,2,balanced,21.65175537109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,1,2,balanced,4.763810424804687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,1,2,power_law_1.01,0.9006310272216798
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,1,2,power_law_1.01,0.943509750366211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,1,2,power_law_1.01,0.9571097564697265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,1,2,power_law_1.01,0.9725734710693359
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,1,2,power_law_1.01,0.9955123138427734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,1,2,balanced,7.347347412109374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,1,2,power_law_1.01,0.5326335906982422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,1,2,power_law_1.01,1.0049612426757812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,1,2,power_law_1.01,1.024501724243164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,1,2,power_law_1.01,1.0639949035644531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,1,2,power_law_1.01,1.1059212493896484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,1,2,power_law_1.01,1.3274751281738282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,1,2,power_law_1.01,1.4909388732910156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,1,2,power_law_1.01,1.6102041625976562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,1,2,power_law_1.01,1.7352499389648437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,1,2,balanced,9.854544677734376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,1,2,power_law_1.01,2.2071192932128905
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,1,2,power_law_1.01,2.8650393676757813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,1,2,power_law_1.01,4.071000366210937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,1,2,balanced,33.57027587890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,1,2,power_law_1.01,5.517840576171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,2,1,power_law_1.2,5.541692504882812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,1,2,power_law_1.01,6.558463745117187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,1,2,power_law_1.01,0.16077184677124023
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,1,2,power_law_1.01,10.0532763671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,1,2,power_law_1.01,0.24175487518310548
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,1,2,power_law_1.01,0.33618175506591796
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,1,2,power_law_1.01,0.3384819030761719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,1,2,power_law_1.01,0.3402969741821289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,1,2,power_law_1.01,0.3433587265014649
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,1,2,power_law_1.01,0.34481536865234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,1,2,power_law_1.01,0.3493900680541992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,1,2,power_law_1.01,14.487947998046874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,1,2,power_law_1.01,0.35000320434570314
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,1,2,power_law_1.01,0.35768192291259765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,1,2,power_law_1.01,0.36118526458740235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,1,2,power_law_1.01,0.24068479537963866
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,1,2,power_law_1.01,0.371987190246582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,1,2,power_law_1.01,0.4082803344726562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,1,2,power_law_1.01,0.4736307144165039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,1,2,power_law_1.01,0.48353408813476567
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,1,2,power_law_1.01,0.5530611038208008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,1,2,power_law_1.01,0.734513931274414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,1,2,balanced,46.2559326171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,1,2,power_law_1.01,0.7642355346679688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,1,2,power_law_1.01,1.3734591674804686
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,1,2,power_law_1.01,1.683494415283203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,1,2,power_law_1.01,3.170626525878906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,1,2,power_law_1.01,23.263254394531252
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,1,2,power_law_1.01,4.900432739257813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,1,2,power_law_1.01,2.4041343688964845
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,1,2,power_law_1.2,0.289172477722168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,1,2,power_law_1.2,0.3867123031616211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,1,2,power_law_1.2,0.5069811248779297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,1,2,power_law_1.2,0.899825897216797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,1,2,power_law_1.2,0.9390681457519531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,1,2,power_law_1.2,0.9561843109130859
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,1,2,power_law_1.2,0.9755161285400391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,1,2,power_law_1.01,6.445972290039063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,1,2,power_law_1.2,0.9940966033935548
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,1,2,power_law_1.2,1.004559326171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,1,2,power_law_1.2,1.0280691528320314
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,1,2,power_law_1.2,1.0634073638916015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,1,2,power_law_1.2,1.106736602783203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,1,2,power_law_1.01,33.42842041015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,1,2,power_law_1.2,1.3074227905273437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,1,2,power_law_1.2,1.52674560546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,1,2,power_law_1.2,1.6422566223144532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,1,2,power_law_1.2,1.7056755065917968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,1,2,power_law_1.2,2.1852096557617187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,1,2,power_law_1.2,2.810093994140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,1,2,power_law_1.2,3.8357312011718747
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,1,2,power_law_1.2,5.57873046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,1,2,power_law_1.01,10.279355468750001
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,1,2,power_law_1.2,6.1796826171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,1,2,power_law_1.01,42.7989892578125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,1,2,power_law_1.01,14.655936279296876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,1,2,power_law_1.2,0.13532159805297853
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,1,2,power_law_1.2,0.20160255432128907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,1,2,power_law_1.2,9.470882568359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,1,2,power_law_1.2,0.32742401123046877
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,1,2,power_law_1.2,0.33899520874023437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,1,2,power_law_1.2,0.34208255767822265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,1,2,power_law_1.2,0.35514240264892577
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,1,2,power_law_1.2,0.35477504730224607
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,1,2,power_law_1.2,0.3502182388305664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,1,2,power_law_1.2,0.3506687927246094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,1,2,power_law_1.2,0.3551206588745117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,1,2,power_law_1.2,0.36294654846191404
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,1,2,power_law_1.2,0.37707263946533204
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,1,2,power_law_1.2,0.24146944046020508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,1,2,power_law_1.2,0.40901889801025393
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,1,2,power_law_1.2,15.43470703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,1,2,power_law_1.2,0.4935884857177735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,1,2,power_law_1.2,0.5132851028442383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,1,2,power_law_1.2,0.570967025756836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,1,2,power_law_1.2,0.7222950744628907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,1,2,power_law_1.2,0.8859506988525391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,1,2,power_law_1.2,1.4503756713867186
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,1,2,power_law_1.2,19.148381347656247
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,1,2,power_law_1.2,1.61357177734375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,1,2,power_law_1.2,2.6348147583007813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,1,2,power_law_1.2,3.5615130615234376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,1,2,power_law_1.2,4.345722961425781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,2,1,balanced,0.1875334358215332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,2,1,balanced,0.36696319580078124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,2,1,balanced,0.6197516632080078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,2,1,balanced,0.6198361587524415
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,2,1,balanced,0.9092339324951173
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,2,1,balanced,0.9142803192138672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,1,2,power_law_1.01,66.9291064453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,2,1,balanced,0.939876480102539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,2,1,balanced,0.9462457275390624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,2,1,balanced,0.978740463256836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,2,1,balanced,0.9883929443359374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,2,1,balanced,1.0032755279541017
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,2,1,balanced,1.0362086486816406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,2,1,balanced,1.0591712188720703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,1,2,power_law_1.2,6.893423461914063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,2,1,balanced,1.1053324890136718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,1,2,power_law_1.2,32.318989257812504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,2,1,balanced,1.38474365234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,2,1,balanced,1.3488255310058594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,2,1,balanced,1.2971757507324218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,2,1,balanced,1.9396569824218752
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,2,1,balanced,2.3603916931152344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,2,1,balanced,3.8087884521484376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,2,1,balanced,4.4107186889648435
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,2,1,balanced,6.992518920898438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,1,2,power_law_1.2,16.929630126953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,1,2,power_law_1.2,41.010810546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,1,2,power_law_1.2,10.993651123046876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,2,1,balanced,9.059842529296876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,2,1,balanced,13.440629882812498
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,2,1,balanced,0.18801984786987305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,2,1,balanced,0.32188865661621097
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,2,1,balanced,0.32169151306152344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,1,2,power_law_1.2,56.040908203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,2,1,balanced,0.10691264152526855
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,2,1,balanced,0.3196121597290039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,2,1,balanced,0.3164473533630371
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,2,1,balanced,0.3217843246459961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,2,1,balanced,18.69726806640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,2,1,balanced,0.3332416152954102
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,2,1,balanced,0.32226112365722653
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,2,1,balanced,0.32537857055664066
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,2,1,balanced,0.3264339065551758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,2,1,balanced,0.3279129409790039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,2,1,balanced,0.5602592086791992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,2,1,balanced,0.6069107055664062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,2,1,balanced,0.33731712341308595
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,2,1,balanced,1.1314157104492186
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,2,1,balanced,1.7084806823730467
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,2,1,balanced,0.370843505859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,2,1,balanced,0.3768966293334961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,2,1,balanced,0.3848812866210937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,2,1,balanced,3.376162414550781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,2,1,balanced,29.64016845703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,2,1,power_law_1.01,0.18179391860961913
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,2,1,balanced,0.39911167144775395
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,2,1,balanced,4.470799865722656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,2,1,power_law_1.01,0.34314048767089844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,2,1,power_law_1.01,0.39666816711425784
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,2,1,balanced,0.8521727752685546
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,2,1,power_law_1.01,0.9079347229003906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,2,1,power_law_1.01,0.9218073272705078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,2,1,power_law_1.01,0.9593843078613281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,2,1,power_law_1.01,0.9504275512695312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,2,1,balanced,6.648028564453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,2,1,power_law_1.01,1.022311019897461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,2,1,power_law_1.01,1.072586212158203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,2,1,power_law_1.01,0.9632313537597657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,2,1,power_law_1.01,0.6088691329956054
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,2,1,power_law_1.01,1.0737312316894532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,2,1,power_law_1.01,1.1971116638183594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,2,1,power_law_1.01,1.2450297546386717
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,2,1,power_law_1.01,1.2748601531982422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,2,1,power_law_1.01,1.3414700317382813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,2,1,power_law_1.01,1.6250726318359376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,2,1,balanced,2.2373670959472656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,2,1,power_law_1.01,2.1193446350097656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,2,1,balanced,9.337299194335937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,2,1,power_law_1.01,2.6474462890624997
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,2,1,power_law_1.01,3.737962341308594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,2,1,power_law_1.01,4.818467712402343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,2,1,balanced,42.5229443359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,2,1,power_law_1.01,7.023624877929687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,2,1,power_law_1.01,9.43815185546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,2,1,power_law_1.01,0.1796499252319336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,2,1,power_law_1.01,0.10647359848022461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,2,1,power_law_1.01,0.20985343933105466
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,2,1,power_law_1.01,0.3208307266235352
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,2,1,power_law_1.01,0.3180524826049805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,2,1,power_law_1.01,0.31930879592895506
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,2,1,power_law_1.01,0.31286399841308593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,2,1,power_law_1.01,0.35889663696289065
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,2,1,power_law_1.01,0.3174080085754395
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,2,1,power_law_1.01,0.358328971862793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,2,1,power_law_1.01,13.9444580078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,2,1,power_law_1.01,0.32240447998046873
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,2,1,power_law_1.01,0.44937919616699223
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,2,1,power_law_1.01,0.3304800033569336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,2,1,power_law_1.01,0.47144256591796874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,2,1,power_law_1.01,0.6235859298706055
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,2,1,power_law_1.01,0.7515167999267578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,2,1,power_law_1.01,0.36792510986328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,2,1,power_law_1.01,1.312442169189453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,2,1,power_law_1.01,18.353304443359374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,2,1,power_law_1.01,0.3942995071411133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,2,1,power_law_1.01,1.8936390686035156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,2,1,power_law_1.01,0.4036985778808594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,2,1,power_law_1.01,2.450746307373047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,2,1,power_law_1.01,1.0485676574707032
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,2,1,power_law_1.2,0.3310707092285156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,2,1,power_law_1.2,0.18016960144042968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,2,1,power_law_1.2,0.38588863372802734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,2,1,power_law_1.01,3.5317459106445312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,2,1,power_law_1.2,0.8572396850585937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,2,1,power_law_1.2,0.9302156829833985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,2,1,power_law_1.2,0.9506285095214844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,2,1,power_law_1.2,0.94697021484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,2,1,power_law_1.2,0.6099609756469727
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,2,1,power_law_1.2,0.9926700592041016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,2,1,power_law_1.2,1.0265440368652343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,2,1,power_law_1.2,1.0706233978271484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,2,1,power_law_1.2,1.0788070678710937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,2,1,power_law_1.01,4.574376831054687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,2,1,power_law_1.2,1.1163507080078126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,2,1,power_law_1.2,1.2526220703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,2,1,power_law_1.2,1.285332489013672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,2,1,power_law_1.2,1.3521318054199218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,2,1,power_law_1.2,1.628515167236328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,2,1,power_law_1.01,28.74514892578125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,2,1,power_law_1.2,2.1394808959960936
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,2,1,power_law_1.2,2.660834655761719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,2,1,power_law_1.01,6.845486450195312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,2,1,power_law_1.2,3.7131698608398436
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,2,1,power_law_1.2,4.847171936035156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,2,1,power_law_1.2,7.013823852539062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,2,1,power_law_1.01,9.208684082031251
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,2,1,power_law_1.2,0.17774463653564454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,2,1,power_law_1.2,9.3907763671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,2,1,power_law_1.01,40.08497802734375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,2,1,power_law_1.2,0.19122495651245117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,2,1,power_law_1.2,0.3198739242553711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,2,1,power_law_1.2,0.32079296112060546
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,2,1,power_law_1.2,0.10643839836120605
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,2,1,power_law_1.2,0.31790464401245117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,2,1,power_law_1.2,0.33085697174072265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,2,1,power_law_1.2,0.35955841064453126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,2,1,power_law_1.2,0.36822334289550784
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,2,1,power_law_1.2,0.3135660743713379
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,2,1,power_law_1.2,0.3626528167724609
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,2,1,power_law_1.2,0.3975718307495117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,2,1,power_law_1.2,0.3078745651245117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,2,1,power_law_1.2,0.414332160949707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,2,1,power_law_1.2,0.46606975555419916
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,2,1,power_law_1.2,0.4732793426513672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,2,1,power_law_1.2,0.5735340881347656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,2,1,power_law_1.2,0.6850227355957031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,2,1,power_law_1.2,13.892843017578125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,2,1,power_law_1.2,0.325048942565918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,2,1,power_law_1.2,1.2372831726074218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,2,1,power_law_1.2,1.8857420349121092
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,2,1,power_law_1.2,2.308687286376953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,2,1,power_law_1.2,1.0039379119873046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,1,2,balanced,0.11523712158203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,1,2,balanced,0.20871679306030275
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,1,2,balanced,0.42285694122314454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,1,2,balanced,0.6977523040771485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,2,1,power_law_1.2,18.615054931640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,1,2,balanced,1.3340467834472656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,1,2,balanced,2.6979275512695313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,2,1,power_law_1.2,3.568161315917969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,1,2,balanced,2.635811767578125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,1,2,balanced,2.666400146484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,1,2,balanced,2.6477017211914062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,1,2,balanced,2.6629556274414066
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,1,2,balanced,2.686473083496094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,1,2,balanced,4.110604858398437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,1,2,balanced,4.134081420898438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,1,2,balanced,4.158485717773438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,2,1,power_law_1.2,4.590453186035156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,1,2,balanced,4.231525268554687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,1,2,balanced,4.271416320800781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,1,2,balanced,4.332105102539062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,1,2,balanced,4.551445617675781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,1,2,balanced,4.750059509277344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,1,2,balanced,5.090927429199219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,1,2,balanced,5.344710693359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,1,2,balanced,6.110596923828124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,2,1,power_law_1.2,29.330644531249998
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,1,2,balanced,6.544887084960938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,1,2,balanced,0.07232128143310547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,1,2,balanced,0.11800704002380372
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,1,2,balanced,0.2125094413757324
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,1,2,balanced,0.39449344635009764
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,1,2,balanced,9.717695922851563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,1,2,balanced,0.6804825592041015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,2,1,power_law_1.2,6.77509521484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,1,2,balanced,1.3222373962402343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,1,2,balanced,1.3145907592773436
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,1,2,balanced,1.3330738830566407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,1,2,balanced,1.3211955261230468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,1,2,balanced,11.928255615234374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,1,2,balanced,1.334281005859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,1,2,balanced,1.34534912109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,1,2,balanced,1.4869580078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,1,2,balanced,1.4669349670410157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,1,2,balanced,1.4691494750976564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,1,2,balanced,1.4829260253906251
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,1,2,balanced,1.489044494628906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,1,2,balanced,1.5233369445800782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,1,2,balanced,1.572026824951172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,1,2,balanced,1.6440013122558592
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,1,2,balanced,1.725727996826172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,1,2,balanced,1.9081024169921874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,1,2,balanced,2.0517132568359377
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,1,2,balanced,2.32974853515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,1,2,balanced,18.742718505859376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,1,2,balanced,3.286398620605469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,1,2,power_law_1.01,0.20471935272216796
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,1,2,power_law_1.01,0.4514393615722656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,1,2,balanced,3.833436279296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,1,2,power_law_1.01,0.5969356918334962
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,1,2,power_law_1.01,0.8481574249267577
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,1,2,power_law_1.01,1.2593433380126953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,1,2,balanced,5.2812915039062505
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,1,2,power_law_1.01,1.4679795837402343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,1,2,balanced,6.811498413085937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,1,2,balanced,23.7655712890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,1,2,power_law_1.01,2.18250244140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,2,1,power_law_1.2,9.220156249999999
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,1,2,power_law_1.01,2.0091416931152346
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,1,2,power_law_1.01,2.0875456237792966
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,1,2,power_law_1.01,3.529841918945313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,1,2,power_law_1.01,3.58450439453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,1,2,power_law_1.01,2.2141273498535154
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,1,2,power_law_1.01,2.2678848266601563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,1,2,power_law_1.01,3.888551635742188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,1,2,power_law_1.01,4.048819274902344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,1,2,power_law_1.01,4.166882629394531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,2,1,power_law_1.2,40.5921240234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,1,2,power_law_1.01,4.433793334960937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,1,2,power_law_1.01,3.6106842041015623
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,1,2,power_law_1.01,5.375574951171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,1,2,power_law_1.01,4.693497619628906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,1,2,power_law_1.01,5.929586181640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,1,2,power_law_1.01,0.212936954498291
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,1,2,power_law_1.01,0.3356377410888672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,1,2,power_law_1.01,7.089971313476562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,1,2,power_law_1.01,0.4425843048095703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,1,2,power_law_1.01,0.6515110778808594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,1,2,power_law_1.01,8.6430859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,1,2,power_law_1.01,0.7314150238037109
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,1,2,power_law_1.01,1.0340364837646485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,1,2,power_law_1.01,0.11763199806213379
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,1,2,power_law_1.01,11.13900634765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,1,2,power_law_1.01,0.9886259460449219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,1,2,power_law_1.01,1.4767974853515624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,1,2,power_law_1.01,1.0452006530761717
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,1,2,power_law_1.01,1.2291661071777344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,1,2,power_law_1.01,14.744090576171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,1,2,power_law_1.01,1.2888230895996093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,1,2,power_law_1.01,1.2785075378417967
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,1,2,power_law_1.01,1.1048076629638672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,1,2,power_law_1.01,1.3087142944335937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,1,2,power_law_1.01,19.213690185546874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,1,2,power_law_1.01,1.383836212158203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,1,2,power_law_1.01,1.486334686279297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,1,2,power_law_1.01,1.5559487915039063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,1,2,power_law_1.01,1.1316582489013673
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,1,2,power_law_1.01,1.9003904724121092
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,1,2,power_law_1.01,2.3908953857421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,1,2,power_law_1.01,2.7515521240234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,1,2,power_law_1.2,0.20687360763549806
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,1,2,power_law_1.2,0.41600383758544923
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,1,2,power_law_1.2,0.5005055999755859
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,1,2,power_law_1.01,3.6841790771484377
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,1,2,power_law_1.2,0.784637451171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,1,2,power_law_1.01,4.514654846191407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,1,2,power_law_1.2,1.1357247924804688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,1,2,power_law_1.2,1.2997056579589843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,1,2,power_law_1.2,2.0788236999511716
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,1,2,power_law_1.2,1.9955545043945313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,1,2,power_law_1.01,25.41904296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,1,2,power_law_1.01,6.313054809570312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,1,2,power_law_1.2,2.135257568359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,1,2,power_law_1.2,2.130240020751953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,1,2,power_law_1.2,2.1677159118652343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,1,2,power_law_1.01,7.989459228515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,1,2,power_law_1.2,3.4514712524414066
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,1,2,power_law_1.2,3.5656268310546877
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,1,2,power_law_1.2,3.6900274658203123
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,1,2,power_law_1.01,1.6735372924804686
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,1,2,power_law_1.2,3.993367004394531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,1,2,power_law_1.2,4.129413146972657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,1,2,power_law_1.2,4.273320922851562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,1,2,power_law_1.2,4.596518249511719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,1,2,power_law_1.2,4.754462585449219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,1,2,power_law_1.2,5.508630981445313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,1,2,power_law_1.2,0.11805439949035645
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,1,2,power_law_1.2,6.152666625976563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,1,2,power_law_1.2,0.20475135803222658
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,1,2,power_law_1.2,0.2876070404052734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,1,2,power_law_1.2,7.263323974609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,1,2,power_law_1.2,0.4509939193725586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,1,2,power_law_1.2,0.5682880020141601
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,1,2,power_law_1.2,0.7269184112548828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,1,2,power_law_1.2,0.9668287658691407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,1,2,power_law_1.2,8.912906494140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,1,2,power_law_1.2,0.9937651062011719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,1,2,power_law_1.2,1.0508006286621092
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,1,2,power_law_1.2,1.0610342407226563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,1,2,power_law_1.2,1.073712615966797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,1,2,power_law_1.2,1.2129843139648437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,1,2,power_law_1.2,1.2451366424560546
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,1,2,power_law_1.2,1.2860287475585939
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,1,2,power_law_1.2,1.3445465087890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,1,2,power_law_1.2,1.38679931640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,1,2,power_law_1.2,12.537198486328126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,1,2,power_law_1.2,1.4458546447753906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,1,2,power_law_1.2,1.5248460388183593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,1,2,power_law_1.2,1.6682623291015624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,1,2,power_law_1.2,1.8210330200195313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,1,2,power_law_1.2,2.0030834960937502
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,1,2,power_law_1.2,14.697486572265626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,1,2,power_law_1.2,2.430234832763672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,1,2,power_law_1.2,2.9444427490234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,2,1,balanced,0.14050368309020994
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,2,1,balanced,0.1940608024597168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,2,1,balanced,0.41467838287353515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,1,2,power_law_1.2,3.9564544677734377
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,2,1,balanced,0.6973478698730469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,1,2,power_law_1.2,4.679514770507812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,2,1,balanced,1.3323770141601563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,1,2,power_law_1.2,6.594086303710938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,1,2,power_law_1.2,21.252309570312498
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,2,1,balanced,2.6910528564453124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,2,1,balanced,2.649020080566406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,2,1,balanced,2.6617752075195313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,1,2,power_law_1.2,8.20253173828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,2,1,balanced,2.6643936157226564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,2,1,balanced,2.701218566894531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,2,1,balanced,2.741533508300781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,2,1,balanced,2.7542272949218747
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,2,1,balanced,2.7917266845703126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,2,1,balanced,2.743740234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,1,2,power_law_1.2,27.173349609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,2,1,balanced,4.1541439819335935
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,2,1,balanced,4.119160461425781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,2,1,balanced,4.219011840820313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,2,1,balanced,4.412983093261719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,2,1,balanced,4.667776489257813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,2,1,balanced,5.035266418457032
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,2,1,balanced,5.296060791015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,2,1,balanced,0.07943039894104004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,2,1,balanced,0.12748991966247558
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,2,1,balanced,0.2197107124328613
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,2,1,balanced,0.3805619049072265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,2,1,balanced,0.6816390228271485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,2,1,balanced,6.224649047851562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,2,1,balanced,6.537301025390624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,2,1,balanced,1.336927947998047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,2,1,balanced,1.3430841064453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,2,1,balanced,1.3505401611328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,2,1,balanced,1.3626348876953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,2,1,balanced,1.3532383728027344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,2,1,balanced,1.3729721069335938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,2,1,balanced,9.634072265624999
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,2,1,balanced,1.3817164611816406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,2,1,balanced,1.3961126708984375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,2,1,balanced,1.3950015258789061
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,2,1,balanced,1.5879411315917968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,2,1,balanced,1.5735494995117187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,2,1,balanced,1.6046424865722657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,2,1,balanced,1.7033357238769533
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,2,1,balanced,1.6585279846191405
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,2,1,balanced,11.765128173828126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,2,1,balanced,1.79740478515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,2,1,balanced,1.8728025817871092
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,2,1,balanced,2.1212780761718752
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,2,1,power_law_1.01,0.1282483196258545
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,2,1,balanced,2.2855130004882813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,2,1,power_law_1.01,0.19291776657104492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,2,1,power_law_1.01,0.4267212677001953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,2,1,power_law_1.01,0.7002265930175782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,2,1,balanced,3.5040280151367185
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,2,1,power_law_1.01,1.038116455078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,2,1,balanced,3.902530517578125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,2,1,power_law_1.01,1.355410614013672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,2,1,power_law_1.01,1.9214994812011716
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,2,1,balanced,5.6935986328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,2,1,power_law_1.01,2.0252915954589845
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,2,1,balanced,17.501470947265624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,2,1,power_law_1.01,2.0762841796875002
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,2,1,power_law_1.01,2.0999058532714843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,2,1,balanced,6.986114501953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,2,1,power_law_1.01,2.210443572998047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,2,1,power_law_1.01,2.326189422607422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,2,1,power_law_1.01,2.3875340270996093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,2,1,power_law_1.01,2.545107879638672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,2,1,power_law_1.01,3.621004028320313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,2,1,power_law_1.01,3.681204528808594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,2,1,power_law_1.01,3.921678161621094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,2,1,balanced,23.1978173828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,2,1,power_law_1.01,4.245377197265625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,2,1,power_law_1.01,4.582212524414063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,2,1,power_law_1.01,5.173676757812499
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,2,1,power_law_1.01,0.08233216285705566
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,2,1,power_law_1.01,5.616381225585938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,2,1,power_law_1.01,0.1269215965270996
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,2,1,power_law_1.01,0.22022911071777346
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,2,1,power_law_1.01,0.38076736450195314
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,2,1,power_law_1.01,0.5482732772827148
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,2,1,power_law_1.01,6.638526000976563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,2,1,power_law_1.01,0.6938111877441406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,2,1,power_law_1.01,0.975766372680664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,2,1,power_law_1.01,1.0325062561035157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,2,1,power_law_1.01,7.835450439453124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,2,1,power_law_1.01,1.0540876770019532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,2,1,power_law_1.01,1.0670329284667968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,2,1,power_law_1.01,1.1199110412597657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,2,1,power_law_1.01,1.1665766143798828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,2,1,power_law_1.01,1.1731423950195312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,2,1,power_law_1.01,1.1906221008300781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,2,1,power_law_1.01,1.3763853454589845
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,2,1,power_law_1.01,1.391320343017578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,2,1,power_law_1.01,10.51163818359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,2,1,power_law_1.01,1.4326675415039063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,2,1,power_law_1.01,1.5101708984375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,2,1,power_law_1.01,1.6225421142578125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,2,1,power_law_1.01,1.7973228454589845
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,2,1,power_law_1.01,13.00541259765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,2,1,power_law_1.01,1.921784362792969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,2,1,power_law_1.01,2.330146484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,2,1,power_law_1.2,0.12647616386413574
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,2,1,power_law_1.01,2.73268798828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,2,1,power_law_1.2,0.1935686492919922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,2,1,power_law_1.2,0.42890560150146484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,2,1,power_law_1.2,0.6598726654052735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,2,1,power_law_1.01,3.589004821777344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,2,1,power_law_1.2,0.9525567626953124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,2,1,power_law_1.01,4.366205444335938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,2,1,power_law_1.2,1.2707039642333986
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,2,1,power_law_1.01,18.03748291015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,2,1,power_law_1.2,1.8905235290527345
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,2,1,power_law_1.01,6.104970703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,2,1,power_law_1.2,2.0325709533691407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,2,1,power_law_1.2,2.1167648315429686
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,2,1,power_law_1.2,1.9819891357421873
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,2,1,power_law_1.2,2.1919033813476565
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,2,1,power_law_1.2,2.360307159423828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,2,1,power_law_1.01,7.781300659179688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,2,1,power_law_1.2,2.4864537048339845
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,2,1,power_law_1.2,2.780540771484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,2,1,power_law_1.2,3.6923046874999996
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,2,1,power_law_1.2,3.811341552734375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,2,1,power_law_1.01,23.4500244140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,2,1,power_law_1.2,3.9867340087890626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,2,1,power_law_1.2,4.396250915527344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,2,1,power_law_1.2,4.664979248046874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,2,1,power_law_1.2,5.2870721435546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,2,1,power_law_1.2,0.08029824256896972
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,2,1,power_law_1.2,5.741610717773438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,2,1,power_law_1.2,0.12786815643310548
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,2,1,power_law_1.2,0.2204096031188965
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,2,1,power_law_1.2,0.35598655700683596
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,2,1,power_law_1.2,0.5030745697021485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,2,1,power_law_1.2,0.657364501953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,2,1,power_law_1.2,6.87555419921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,2,1,power_law_1.2,0.9551660919189453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,2,1,power_law_1.2,1.0147154998779295
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,2,1,power_law_1.2,8.246044921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,2,1,power_law_1.2,1.0386758422851563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,2,1,power_law_1.2,1.073749771118164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,2,1,power_law_1.2,1.1085369873046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,2,1,power_law_1.2,1.1507628631591795
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,2,1,power_law_1.2,1.1766655731201172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,2,1,power_law_1.2,1.2166611480712892
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,2,1,power_law_1.2,10.722406005859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,2,1,power_law_1.2,1.4005119323730468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,2,1,power_law_1.2,1.4252217102050782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,2,1,power_law_1.2,1.476275177001953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,2,1,power_law_1.2,1.5536480712890626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,2,1,power_law_1.2,1.6983526611328124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,2,1,power_law_1.2,1.9021452331542967
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,2,1,power_law_1.2,13.183023681640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,2,1,power_law_1.2,2.0337753295898438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,2,1,power_law_1.2,2.4701023864746094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,1,2,balanced,0.10784895896911621
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,1,2,balanced,0.16137088775634764
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,2,1,power_law_1.2,2.8805529785156248
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,1,2,balanced,0.34780799865722656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,1,2,balanced,0.6036006546020507
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,1,2,balanced,1.1407129669189453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,2,1,power_law_1.2,3.6642337036132817
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,1,2,balanced,2.295353546142578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,2,1,power_law_1.2,4.617575073242188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,1,2,balanced,2.258784027099609
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,1,2,balanced,2.290496063232422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,2,1,power_law_1.2,18.402286376953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,1,2,balanced,2.31399169921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,2,1,power_law_1.2,6.219627685546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,1,2,balanced,2.2916441345214844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,1,2,balanced,2.3250521850585937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,1,2,balanced,3.5568319702148434
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,2,1,power_law_1.2,7.930609130859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,1,2,balanced,3.580296936035156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,1,2,balanced,3.613352966308594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,1,2,balanced,3.671430358886719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,1,2,balanced,3.7099493408203124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,1,2,balanced,3.7699685668945313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,1,2,balanced,3.933968505859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,2,1,power_law_1.2,23.465383300781248
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,1,2,balanced,4.104104919433594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,1,2,balanced,4.356683654785156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,1,2,balanced,0.06424575805664062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,1,2,balanced,4.545859985351562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,1,2,balanced,0.10514944076538087
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,1,2,balanced,0.1846950340270996
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,1,2,balanced,5.106519165039062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,1,2,balanced,0.34024318695068356
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,1,2,balanced,0.5871219253540039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,1,2,balanced,5.541246948242188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,1,2,balanced,1.1366668701171876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,1,2,balanced,1.1380134582519532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,1,2,balanced,1.139283218383789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,1,2,balanced,1.1435404968261718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,1,2,balanced,1.1508902740478515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,1,2,balanced,8.253634643554687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,1,2,balanced,1.160517120361328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,1,2,balanced,1.2915367126464843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,1,2,balanced,1.2474854278564453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,1,2,balanced,1.3008242797851561
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,1,2,balanced,1.2677862548828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,1,2,balanced,1.2687219238281249
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,1,2,balanced,1.325269775390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,1,2,balanced,1.3594560241699218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,1,2,balanced,1.4573068237304687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,1,2,balanced,1.4864242553710938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,1,2,balanced,10.345653076171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,1,2,balanced,1.5850355529785156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,1,2,balanced,1.7625447082519532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,1,2,balanced,1.970713653564453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,1,2,power_law_1.01,0.15764736175537108
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,1,2,power_law_1.01,0.35930240631103516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,1,2,power_law_1.01,0.4801523208618164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,1,2,balanced,2.942318115234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,1,2,power_law_1.01,0.7151423645019531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,1,2,balanced,3.3596621704101564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,1,2,power_law_1.01,0.996209945678711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,1,2,power_law_1.01,1.251687698364258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,1,2,balanced,4.585862426757812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,1,2,power_law_1.01,1.7195097351074218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,1,2,balanced,15.813074951171876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,1,2,power_law_1.01,1.7230873107910156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,1,2,power_law_1.01,1.755449523925781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,1,2,power_law_1.01,1.7805850219726562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,1,2,balanced,5.876312255859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,1,2,power_law_1.01,1.9259852600097656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,1,2,power_law_1.01,2.9407333374023437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,1,2,power_law_1.01,3.0688320922851564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,1,2,power_law_1.01,3.2427136230468747
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,1,2,power_law_1.01,3.2715570068359376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,1,2,balanced,20.066295166015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,1,2,power_law_1.01,3.600001220703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,1,2,power_law_1.01,3.8398092651367186
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,1,2,power_law_1.01,4.039658203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,1,2,power_law_1.01,3.1373272705078126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,1,2,power_law_1.01,4.957894287109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,1,2,power_law_1.01,0.10542464256286621
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,1,2,power_law_1.01,0.1844710350036621
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,1,2,power_law_1.01,5.964364624023437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,1,2,power_law_1.01,0.26959232330322264
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,1,2,power_law_1.01,0.4043865585327149
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,1,2,power_law_1.01,0.495307502746582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,1,2,power_law_1.01,4.496711730957031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,1,2,power_law_1.01,7.040970458984376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,1,2,power_law_1.01,0.636794891357422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,1,2,power_law_1.01,0.8320947265624999
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,1,2,power_law_1.01,0.8336844635009765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,1,2,power_law_1.01,0.8897305297851563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,1,2,power_law_1.01,0.9597555541992188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,1,2,power_law_1.01,1.0573004913330077
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,1,2,power_law_1.01,1.0769471740722656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,1,2,power_law_1.01,10.284829101562499
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,1,2,power_law_1.01,1.120943374633789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,1,2,power_law_1.01,1.1778880310058593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,1,2,power_law_1.01,11.796375732421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,1,2,power_law_1.01,1.3182424926757812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,1,2,power_law_1.01,1.6246707153320312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,1,2,power_law_1.01,0.9167628479003905
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,1,2,power_law_1.01,2.0774873352050784
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,1,2,power_law_1.01,16.799202880859376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,1,2,power_law_1.01,3.197699890136719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,1,2,power_law_1.01,1.0874111938476563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,1,2,power_law_1.01,4.035802917480469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,1,2,power_law_1.01,1.1502642822265625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,1,2,power_law_1.2,0.1635865592956543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,1,2,power_law_1.01,21.89365478515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,1,2,power_law_1.2,0.3499763107299805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,1,2,power_law_1.2,0.4453030395507812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,1,2,power_law_1.01,1.2444403076171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,1,2,power_law_1.2,0.671435546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,1,2,power_law_1.2,0.8785395050048829
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,1,2,power_law_1.2,1.1150899505615235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,1,2,power_law_1.01,1.5003904724121093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,1,2,power_law_1.2,1.616944580078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,1,2,power_law_1.2,1.6635903930664064
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,1,2,power_law_1.2,1.7083161926269532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,1,2,power_law_1.01,5.842045288085937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,1,2,power_law_1.2,1.7546867370605468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,1,2,power_law_1.2,1.9684019470214842
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,1,2,power_law_1.2,2.9257907104492187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,1,2,power_law_1.2,3.0874624633789063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,1,2,power_law_1.2,3.1990771484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,1,2,power_law_1.01,2.417973785400391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,1,2,power_law_1.01,6.704517211914063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,1,2,power_law_1.2,3.342351379394531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,1,2,power_law_1.2,3.425838012695313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,1,2,power_law_1.2,3.7025881958007814
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,1,2,power_law_1.2,3.9732427978515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,1,2,power_law_1.2,4.185918579101562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,1,2,power_law_1.2,4.647819519042969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,1,2,power_law_1.2,6.181931762695312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,1,2,power_law_1.2,0.1053222370147705
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,1,2,power_law_1.2,0.18475648880004883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,1,2,power_law_1.2,0.25418495178222655
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,1,2,power_law_1.2,0.37583358764648434
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,1,2,power_law_1.2,5.0720281982421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,1,2,power_law_1.2,0.48637054443359373
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,1,2,power_law_1.2,0.5800307083129883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,1,2,power_law_1.2,0.8198834991455077
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,1,2,power_law_1.2,0.8502553558349609
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,1,2,power_law_1.2,0.8867955017089845
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,1,2,power_law_1.2,10.803031005859376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,1,2,power_law_1.2,0.9305894470214844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,1,2,power_law_1.2,0.9715443420410157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,1,2,power_law_1.2,7.380308227539063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,1,2,power_law_1.2,1.0534989166259767
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,1,2,power_law_1.2,1.0919654083251953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,1,2,power_law_1.2,1.123592987060547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,1,2,power_law_1.2,1.1329177856445312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,1,2,power_law_1.2,1.1702796936035156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,1,2,power_law_1.2,12.383560791015626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,1,2,power_law_1.2,1.2410214233398438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,1,2,power_law_1.2,1.2959129333496093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,1,2,power_law_1.2,1.3902847290039062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,1,2,power_law_1.2,1.546859588623047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,1,2,power_law_1.2,1.731266632080078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,1,2,power_law_1.2,2.1286528015136716
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,2,1,balanced,0.1087155246734619
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,1,2,power_law_1.2,2.623424072265625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,2,1,balanced,0.166125431060791
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,2,1,balanced,0.37020225524902345
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,2,1,balanced,0.6056889724731446
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,1,2,power_law_1.2,3.4030706787109373
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,1,2,power_law_1.2,4.221358032226563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,2,1,balanced,1.143660125732422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,1,2,power_law_1.2,17.54166748046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,2,1,balanced,2.255114288330078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,1,2,power_law_1.2,5.992207641601563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,2,1,balanced,2.277024688720703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,2,1,balanced,2.31018310546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,2,1,balanced,2.3411322021484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,1,2,power_law_1.2,7.34503662109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,2,1,balanced,2.3419871520996094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,2,1,balanced,2.308293762207031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,2,1,balanced,2.3401478576660155
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,2,1,balanced,2.423875274658203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,2,1,balanced,2.4251161193847657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,1,2,power_law_1.2,22.24088623046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,2,1,balanced,3.568966979980469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,2,1,balanced,3.598851318359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,2,1,balanced,3.6554861450195313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,2,1,balanced,3.8680255126953127
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,2,1,balanced,4.009550781250001
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,2,1,balanced,4.2952947998046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,2,1,balanced,0.07261439800262451
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,2,1,balanced,0.11355456352233886
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,2,1,balanced,4.491329956054687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,2,1,balanced,0.19355968475341795
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,2,1,balanced,0.31646400451660156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,2,1,balanced,5.1799645996093755
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,2,1,balanced,0.5914815902709961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,2,1,balanced,1.1571692657470702
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,2,1,balanced,5.6489630126953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,2,1,balanced,1.1741574096679688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,2,1,balanced,1.165494384765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,2,1,balanced,1.169668502807617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,2,1,balanced,1.192449951171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,2,1,balanced,1.195269775390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,2,1,balanced,1.1919814300537108
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,2,1,balanced,1.2056403350830078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,2,1,balanced,1.2206681823730468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,2,1,balanced,1.3722872924804688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,2,1,balanced,1.3751008605957031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,2,1,balanced,8.130082397460937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,2,1,balanced,1.4229830932617187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,2,1,balanced,1.4302854919433594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,2,1,balanced,1.4363020324707032
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,2,1,balanced,10.186407470703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,2,1,balanced,1.5422239685058594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,2,1,balanced,1.5987333679199218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,2,1,balanced,1.8030047607421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,2,1,balanced,1.960298309326172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,2,1,power_law_1.01,0.10452223777770997
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,2,1,power_law_1.01,0.16603584289550782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,2,1,power_law_1.01,0.35827007293701174
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,2,1,balanced,2.9770693969726563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,2,1,balanced,3.2615124511718747
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,2,1,power_law_1.01,0.6032633590698242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,2,1,power_law_1.01,0.8892518615722655
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,2,1,power_law_1.01,1.1514956665039062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,2,1,balanced,4.706437072753906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,2,1,power_law_1.01,1.6455795288085937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,2,1,power_law_1.01,1.7375794982910158
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,2,1,balanced,15.027171630859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,2,1,power_law_1.01,1.7734675598144531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,2,1,balanced,5.875480346679687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,2,1,power_law_1.01,1.7927807617187501
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,2,1,power_law_1.01,1.9064390563964841
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,2,1,power_law_1.01,2.039925079345703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,2,1,power_law_1.01,2.0422393798828127
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,2,1,power_law_1.01,2.2581304931640624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,2,1,power_law_1.01,3.134796142578125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,2,1,balanced,18.3829345703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,2,1,power_law_1.01,3.1859423828124998
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,2,1,power_law_1.01,3.382384033203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,2,1,power_law_1.01,3.635992431640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,2,1,power_law_1.01,3.912336120605469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,2,1,power_law_1.01,4.318910827636719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,2,1,power_law_1.01,0.0737440013885498
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,2,1,power_law_1.01,0.11343744277954101
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,2,1,power_law_1.01,4.729216003417969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,2,1,power_law_1.01,0.19332223892211914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,2,1,power_law_1.01,0.33245761871337887
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,2,1,power_law_1.01,5.491737670898438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,2,1,power_law_1.01,0.47188159942626956
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,2,1,power_law_1.01,0.6055731201171876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,2,1,power_law_1.01,0.842081298828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,2,1,power_law_1.01,0.8891980743408203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,2,1,power_law_1.01,6.630189208984374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,2,1,power_law_1.01,0.9107315063476562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,2,1,power_law_1.01,0.9206534576416014
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,2,1,power_law_1.01,0.970798110961914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,2,1,power_law_1.01,1.002346878051758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,2,1,power_law_1.01,1.0152127838134766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,2,1,power_law_1.01,8.848944091796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,2,1,power_law_1.01,1.1472518157958984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,2,1,power_law_1.01,1.1982937622070313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,2,1,power_law_1.01,1.0266905975341796
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,2,1,power_law_1.01,1.2472262573242188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,2,1,power_law_1.01,1.2922802734375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,2,1,power_law_1.01,1.4240614318847657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,2,1,power_law_1.01,11.222034912109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,2,1,power_law_1.01,1.5529087829589845
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,2,1,power_law_1.01,1.6887974548339844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,2,1,power_law_1.01,2.019194946289063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,2,1,power_law_1.2,0.10484160423278807
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,2,1,power_law_1.01,2.3700685119628906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,2,1,power_law_1.2,0.16564159393310546
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,2,1,power_law_1.2,0.3603033447265625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,2,1,power_law_1.2,0.5653504180908203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,2,1,power_law_1.01,3.2073422241210934
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,2,1,power_law_1.2,0.8226751708984376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,2,1,power_law_1.01,3.7512939453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,2,1,power_law_1.2,1.0890284729003905
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,2,1,power_law_1.2,1.6289318847656251
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,2,1,power_law_1.01,15.45281494140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,2,1,power_law_1.2,1.7011827087402345
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,2,1,power_law_1.2,1.743034210205078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,2,1,power_law_1.2,1.8166188049316407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,2,1,power_law_1.01,5.369157104492187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,2,1,power_law_1.01,6.5101953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,2,1,power_law_1.2,1.9031968688964844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,2,1,power_law_1.2,1.9962483215332032
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,2,1,power_law_1.2,2.052472381591797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,2,1,power_law_1.2,2.3537120056152343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,2,1,power_law_1.01,19.861806640624998
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,2,1,power_law_1.2,3.192011413574219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,2,1,power_law_1.2,3.3003710937500004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,2,1,power_law_1.2,3.436081848144531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,2,1,power_law_1.2,3.742974853515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,2,1,power_law_1.2,3.995888061523438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,2,1,power_law_1.2,4.499647216796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,2,1,power_law_1.2,0.07416704177856445
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,2,1,power_law_1.2,4.800151672363281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,2,1,power_law_1.2,0.11338687896728517
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,2,1,power_law_1.2,0.1961510467529297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,2,1,power_law_1.2,0.30854848861694334
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,2,1,power_law_1.2,0.43532161712646483
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,2,1,power_law_1.2,0.5680755233764648
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,2,1,power_law_1.2,5.8265618896484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,2,1,power_law_1.2,0.824922866821289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,2,1,power_law_1.2,0.8728991699218749
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,2,1,power_law_1.2,6.911292114257813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,2,1,power_law_1.2,0.8980332946777343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,2,1,power_law_1.2,0.9260063934326173
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,2,1,power_law_1.2,0.955472640991211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,2,1,power_law_1.2,0.9918061065673829
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,2,1,power_law_1.2,1.0169036865234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,2,1,power_law_1.2,1.0484998321533203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,2,1,power_law_1.2,9.222582397460936
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,2,1,power_law_1.2,1.1997869110107422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,2,1,power_law_1.2,1.2361644744873046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,2,1,power_law_1.2,1.2869920349121093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,2,1,power_law_1.2,1.3661964416503907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,2,1,power_law_1.2,1.4315423583984375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,2,1,power_law_1.2,1.6211878967285156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,2,1,power_law_1.2,11.2819970703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,2,1,power_law_1.2,2.1414842224121093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,2,1,power_law_1.2,1.790862121582031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,1,2,balanced,0.02867072105407715
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,1,2,balanced,0.03804415941238403
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,1,2,balanced,0.054302721023559576
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,1,2,balanced,0.0945139217376709
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,2,1,power_law_1.2,2.4809689331054687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,1,2,balanced,0.15873536109924316
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,1,2,balanced,0.17691392898559571
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,1,2,balanced,0.16207616806030273
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,1,2,balanced,0.166112003326416
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,1,2,balanced,0.21943424224853514
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,1,2,balanced,0.22468608856201172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,2,1,power_law_1.2,3.161335144042969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,1,2,balanced,0.22588672637939453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,1,2,balanced,0.24447359085083006
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,1,2,balanced,0.2342361640930176
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,1,2,balanced,0.2447987174987793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,1,2,balanced,0.24986751556396483
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,1,2,balanced,0.2565567970275879
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,1,2,balanced,0.27116287231445313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,1,2,balanced,0.290437126159668
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,1,2,balanced,0.3086886405944824
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,2,1,power_law_1.2,3.854659729003906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,1,2,balanced,0.37040641784667966
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,1,2,balanced,0.4159564971923828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,1,2,balanced,0.5754150390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,1,2,balanced,0.7362879943847657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,2,1,power_law_1.2,15.62607421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,1,2,balanced,1.0491506958007812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,1,2,balanced,0.02579328060150147
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,1,2,balanced,0.024951040744781494
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,1,2,balanced,0.04060544013977051
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,1,2,balanced,0.06337664127349854
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,2,1,power_law_1.2,5.3328271484375005
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,1,2,balanced,0.10180864334106446
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,1,2,balanced,1.3561331176757814
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,1,2,balanced,0.10134528160095216
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,1,2,balanced,0.10557184219360352
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,1,2,balanced,0.11935872077941893
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,1,2,balanced,0.11645183563232422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,1,2,balanced,0.11688832283020019
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,1,2,balanced,0.11850624084472657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,1,2,balanced,0.11946368217468262
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,1,2,balanced,0.12509823799133302
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,1,2,balanced,0.12024703979492188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,1,2,balanced,0.12301440238952636
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,1,2,balanced,0.12651776313781737
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,1,2,balanced,1.9869900512695313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,1,2,balanced,0.13860735893249512
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,1,2,balanced,0.14198399543762208
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,1,2,balanced,0.16229120254516602
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,1,2,balanced,0.17885440826416016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,1,2,balanced,0.21224576950073243
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,1,2,balanced,0.3011840057373047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,1,2,balanced,0.3458163070678711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,1,2,balanced,2.61125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,1,2,balanced,0.4913625717163086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,1,2,power_law_1.01,0.03757184028625488
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,1,2,balanced,0.6324441528320313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,1,2,power_law_1.01,0.0616819190979004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,1,2,power_law_1.01,0.07800576210021973
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,1,2,power_law_1.01,0.10170623779296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,1,2,balanced,0.924963836669922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,1,2,power_law_1.01,0.14158207893371583
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,1,2,power_law_1.01,0.15046527862548828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,1,2,power_law_1.01,0.15252351760864258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,1,2,balanced,1.2117171478271485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,1,2,power_law_1.01,0.20751359939575192
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,1,2,power_law_1.01,0.21242368698120115
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,1,2,power_law_1.01,0.22070016860961914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,1,2,power_law_1.01,0.22671104431152345
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,1,2,power_law_1.01,0.23024255752563477
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,1,2,power_law_1.01,0.2343641662597656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,2,1,power_law_1.2,6.978982543945312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,1,2,power_law_1.01,0.24279552459716797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,1,2,power_law_1.01,0.264846076965332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,1,2,power_law_1.01,0.30123008728027345
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,1,2,power_law_1.01,0.31437824249267576
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,1,2,power_law_1.01,0.46413822174072267
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,1,2,power_law_1.01,0.4065574264526367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,1,2,power_law_1.01,0.6283238220214844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,1,2,power_law_1.01,0.8084595489501953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,1,2,power_law_1.01,0.2598374366760254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,1,2,power_law_1.01,0.023216640949249266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,1,2,power_law_1.01,1.2138931274414062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,1,2,power_law_1.01,0.03614464044570923
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,1,2,power_law_1.01,0.0484006404876709
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,1,2,power_law_1.01,0.060247039794921874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,1,2,power_law_1.01,0.0649280023574829
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,1,2,power_law_1.01,1.4508262634277345
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,1,2,power_law_1.01,0.09164159774780274
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,2,1,power_law_1.2,20.0930908203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,1,2,power_law_1.01,0.0959385585784912
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,1,2,power_law_1.01,0.11110912322998048
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,1,2,power_law_1.01,0.10858240127563476
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,1,2,power_law_1.01,0.11366016387939452
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,1,2,power_law_1.01,0.04656000137329101
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,1,2,power_law_1.01,0.11731200218200684
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,1,2,power_law_1.01,0.1184102439880371
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,1,2,power_law_1.01,2.2100146484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,1,2,power_law_1.01,0.12248576164245606
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,1,2,power_law_1.01,0.1262335968017578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,1,2,power_law_1.01,0.15164159774780273
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,1,2,power_law_1.01,0.17089151382446288
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,1,2,power_law_1.01,0.2069209671020508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,1,2,power_law_1.01,0.24407039642333986
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,1,2,power_law_1.01,0.31787519454956054
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,1,2,power_law_1.01,0.09727359771728515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,1,2,power_law_1.01,2.8370098876953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,1,2,power_law_1.01,0.39672576904296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,1,2,power_law_1.01,0.12334207534790038
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,1,2,power_law_1.01,0.551863021850586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,1,2,power_law_1.2,0.03735039949417114
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,1,2,power_law_1.2,0.04322944164276123
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,1,2,power_law_1.01,0.6892467498779297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,1,2,power_law_1.01,0.13683584213256836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,1,2,power_law_1.2,0.058721280097961424
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,1,2,power_law_1.2,0.07427584171295167
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,1,2,power_law_1.2,0.10205951690673828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,1,2,power_law_1.2,0.14866559982299804
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,1,2,power_law_1.2,0.1383897590637207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,1,2,power_law_1.2,0.1554700756072998
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,1,2,power_law_1.01,1.0963353729248047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,1,2,power_law_1.2,0.21187711715698243
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,1,2,power_law_1.2,0.22025728225708008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,1,2,power_law_1.2,0.22284416198730467
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,1,2,power_law_1.01,1.3044723510742187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,1,2,power_law_1.2,0.23029375076293945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,1,2,power_law_1.2,0.24593280792236327
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,1,2,power_law_1.2,0.2645747184753418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,1,2,power_law_1.2,0.2683008003234863
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,1,2,power_law_1.2,0.26491775512695315
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,1,2,power_law_1.2,0.3033087921142578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,1,2,power_law_1.2,0.3203622436523438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,1,2,power_law_1.2,0.4051327896118164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,1,2,power_law_1.2,0.4756492614746094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,1,2,power_law_1.2,0.21222272872924805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,1,2,power_law_1.2,0.6574655914306641
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,1,2,power_law_1.2,0.024572160243988037
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,1,2,power_law_1.2,0.03452415943145752
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,1,2,power_law_1.2,0.04505856037139892
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,1,2,power_law_1.2,0.05410687923431397
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,1,2,power_law_1.2,0.0658572816848755
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,1,2,power_law_1.2,0.0900812816619873
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,1,2,power_law_1.2,0.0960319995880127
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,1,2,power_law_1.2,0.09773311614990235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,1,2,power_law_1.2,0.8372415924072266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,1,2,power_law_1.2,0.11066495895385742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,1,2,power_law_1.2,0.11031935691833497
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,1,2,power_law_1.2,0.11294591903686524
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,1,2,power_law_1.2,0.1148966407775879
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,1,2,power_law_1.2,0.11766143798828126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,1,2,power_law_1.2,0.12069120407104492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,1,2,power_law_1.2,1.2784857940673828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,1,2,power_law_1.2,0.12652671813964844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,1,2,power_law_1.2,0.13167615890502932
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,1,2,power_law_1.2,0.13913344383239745
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,1,2,power_law_1.2,0.15318655967712402
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,1,2,power_law_1.2,0.17042047500610352
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,1,2,power_law_1.2,0.20921472549438475
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,1,2,power_law_1.2,0.24539520263671877
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,1,2,power_law_1.2,0.32635776519775395
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,1,2,power_law_1.2,0.41015167236328126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,1,2,power_law_1.2,0.5577497482299805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,1,2,power_law_1.2,1.4968051147460937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,2,1,balanced,0.04158783912658691
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,2,1,balanced,0.03191296100616455
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,1,2,power_law_1.2,1.1359053039550782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,2,1,balanced,0.063503999710083
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,2,1,balanced,0.09564736366271973
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,1,2,power_law_1.2,0.7003392028808594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,1,2,power_law_1.2,2.2034393310546876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,2,1,balanced,0.16822143554687502
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,2,1,balanced,0.16718591690063475
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,2,1,balanced,0.17853504180908203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,2,1,balanced,0.18167808532714844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,2,1,balanced,0.17311296463012696
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,2,1,balanced,0.17894016265869142
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,2,1,balanced,0.18013055801391603
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,2,1,balanced,0.1765939140319824
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,2,1,balanced,0.18050559997558593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,2,1,balanced,0.232174072265625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,2,1,balanced,0.25193344116210936
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,2,1,balanced,0.18699520111083984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,2,1,balanced,0.27061119079589846
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,2,1,balanced,0.3909817504882812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,2,1,balanced,0.435728645324707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,2,1,balanced,0.6228902435302734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,1,2,power_law_1.2,1.3188621520996093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,2,1,balanced,0.18946367263793945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,2,1,balanced,0.8182067108154296
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,2,1,balanced,0.02780159950256348
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,2,1,balanced,0.03207616090774536
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,2,1,balanced,0.05037248134613037
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,2,1,balanced,0.07413824081420899
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,2,1,balanced,0.11470015525817871
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,2,1,balanced,1.1819225311279298
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,2,1,balanced,0.11727487564086914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,2,1,balanced,0.1196115207672119
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,2,1,balanced,0.1193727970123291
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,2,1,balanced,0.19180351257324219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,2,1,balanced,0.12042495727539063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,2,1,balanced,0.12049856185913085
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,2,1,balanced,0.13943936347961425
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,2,1,balanced,0.11517696380615235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,2,1,balanced,0.14479616165161133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,2,1,balanced,0.14574336051940917
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,2,1,balanced,0.14974656105041503
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,2,1,balanced,0.15747903823852538
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,2,1,balanced,0.16675647735595703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,2,1,balanced,0.2044704055786133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,2,1,balanced,0.1844825553894043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,2,1,balanced,0.1413830375671387
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,2,1,balanced,0.20261823654174807
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,1,2,power_law_1.2,2.970830078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,2,1,balanced,0.19664127349853516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,2,1,balanced,0.29701824188232423
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,2,1,balanced,0.33522369384765627
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,2,1,balanced,0.4767871856689453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,2,1,power_law_1.01,0.03211328029632569
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,2,1,balanced,0.6198067092895508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,2,1,power_law_1.01,0.041507840156555176
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,2,1,balanced,0.9036243438720704
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,2,1,power_law_1.01,0.0813644790649414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,2,1,balanced,1.1812614440917968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,2,1,balanced,0.14185471534729005
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,2,1,power_law_1.01,0.16713855743408204
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,2,1,power_law_1.01,0.1782963180541992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,2,1,power_law_1.01,0.06217984199523926
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,2,1,power_law_1.01,0.18468544006347656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,2,1,power_law_1.01,0.18969215393066405
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,2,1,power_law_1.01,0.1943769645690918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,2,1,power_law_1.01,0.14229056358337402
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,2,1,power_law_1.01,0.213953914642334
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,2,1,power_law_1.01,0.20628543853759768
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,2,1,power_law_1.01,0.10118464469909667
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,2,1,power_law_1.01,0.2518796730041504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,2,1,power_law_1.01,0.2973535919189453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,2,1,power_law_1.01,0.34883007049560544
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,2,1,power_law_1.01,0.15868672370910644
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,2,1,power_law_1.01,0.43019905090332033
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,2,1,power_law_1.01,0.16489471435546874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,2,1,power_law_1.01,0.517968635559082
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,2,1,power_law_1.01,0.7087718200683594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,2,1,power_law_1.01,0.17060928344726561
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,2,1,power_law_1.01,0.02830656051635742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,2,1,power_law_1.01,0.03461695909500122
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,2,1,power_law_1.01,0.8993523406982422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,2,1,power_law_1.01,0.049752321243286136
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,2,1,power_law_1.01,0.06747136116027833
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,2,1,power_law_1.01,0.07535168170928955
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,2,1,power_law_1.01,0.10235839843750001
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,2,1,balanced,1.5820672607421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,2,1,power_law_1.01,0.1080185604095459
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,2,1,power_law_1.01,0.11052991867065429
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,2,1,power_law_1.01,0.11357376098632813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,2,1,power_law_1.01,0.1134432029724121
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,2,1,power_law_1.01,0.11686143875122071
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,2,1,power_law_1.01,0.1359379196166992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,2,1,power_law_1.01,0.13754688262939455
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,2,1,power_law_1.01,1.742467803955078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,2,1,power_law_1.01,0.1453388786315918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,2,1,power_law_1.01,0.1489900779724121
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,2,1,power_law_1.01,0.15935935974121093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,2,1,power_law_1.01,0.17366144180297852
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,2,1,power_law_1.01,0.18052480697631837
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,2,1,power_law_1.01,0.21701887130737302
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,2,1,power_law_1.01,0.2530393600463867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,2,1,power_law_1.01,0.18862783432006835
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,2,1,power_law_1.01,0.3217984008789062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,2,1,power_law_1.01,0.1413804817199707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,2,1,power_law_1.01,0.529743995666504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,2,1,power_law_1.01,1.2864959716796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,2,1,power_law_1.01,0.2192505645751953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,2,1,power_law_1.2,0.041728639602661134
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,2,1,power_law_1.01,0.6709817504882812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,2,1,power_law_1.2,0.09557184219360351
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,2,1,power_law_1.01,0.9556473541259767
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,2,1,power_law_1.01,0.38316223144531253
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,2,1,power_law_1.2,0.1603775978088379
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,2,1,power_law_1.01,1.2567040252685548
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,2,1,power_law_1.2,0.17378816604614258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,2,1,power_law_1.2,0.16972799301147462
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,2,1,power_law_1.2,0.18030784606933595
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,2,1,power_law_1.2,0.18666879653930663
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,2,1,power_law_1.2,0.19167423248291016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,2,1,power_law_1.2,0.19718015670776368
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,2,1,power_law_1.2,0.2151468849182129
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,2,1,power_law_1.2,0.2063852882385254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,2,1,power_law_1.2,0.22002687454223632
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,2,1,power_law_1.2,0.25366079330444335
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,2,1,power_law_1.2,0.30143999099731444
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,2,1,power_law_1.2,0.35306304931640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,2,1,power_law_1.2,0.05948031902313232
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,2,1,power_law_1.2,0.07729983806610108
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,2,1,power_law_1.2,0.0320959997177124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,2,1,power_law_1.2,0.43210304260253907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,2,1,power_law_1.2,0.14412223815917968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,2,1,power_law_1.2,0.0280019211769104
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,2,1,power_law_1.2,0.035285120010375974
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,2,1,power_law_1.2,0.5322643280029297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,2,1,power_law_1.2,0.04865087985992432
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,2,1,power_law_1.2,0.06083199977874756
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,2,1,power_law_1.2,0.07237184047698975
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,2,1,power_law_1.2,0.104901123046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,2,1,power_law_1.2,0.1095961570739746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,2,1,power_law_1.2,0.16845695495605467
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,2,1,power_law_1.2,0.11205120086669922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,2,1,power_law_1.2,0.11491711616516113
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,2,1,power_law_1.2,0.11411904335021972
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,2,1,power_law_1.2,0.11742591857910156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,2,1,power_law_1.2,0.13744256019592285
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,2,1,power_law_1.2,0.13850560188293456
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,2,1,power_law_1.2,0.1909600067138672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,2,1,power_law_1.2,0.1433836841583252
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,2,1,power_law_1.2,0.14820480346679688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,2,1,power_law_1.2,0.15236096382141112
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,2,1,power_law_1.2,0.16238143920898437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,2,1,power_law_1.2,0.17117504119873045
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,2,1,power_law_1.2,0.18466047286987303
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,2,1,power_law_1.2,0.22037696838378906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,2,1,power_law_1.2,0.2550464057922363
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,2,1,power_law_1.2,0.7128192138671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,2,1,power_law_1.2,0.3207628631591797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,2,1,power_law_1.2,0.39155647277832034
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,1,2,balanced,0.061304321289062505
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,1,2,balanced,0.08968192100524902
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,1,2,balanced,0.1745689582824707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,1,2,balanced,0.3023756790161133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,1,2,balanced,0.5764608001708984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,1,2,balanced,0.5809100723266601
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,1,2,balanced,0.5825024032592774
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,2,1,power_law_1.2,0.5300064086914062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,1,2,balanced,0.5856729507446289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,1,2,balanced,0.8558233642578126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,2,1,power_law_1.2,0.9068812561035156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,1,2,balanced,0.861875228881836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,1,2,balanced,0.8716722869873047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,1,2,balanced,0.8907647705078124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,1,2,balanced,0.8988172912597656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,1,2,balanced,0.9213145446777343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,1,2,balanced,0.9542182159423828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,1,2,balanced,0.9607884979248047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,1,2,balanced,1.0180902099609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,1,2,balanced,1.0926207733154296
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,1,2,balanced,1.141406707763672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,1,2,balanced,1.306279754638672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,1,2,balanced,1.4481741333007814
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,2,1,power_law_1.2,0.6729682922363281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,1,2,balanced,2.0736614990234377
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,1,2,balanced,2.5916262817382814
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,1,2,balanced,3.846092834472656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,1,2,balanced,0.04174335956573486
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,1,2,balanced,0.058823680877685545
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,1,2,balanced,0.09834239959716798
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,1,2,balanced,0.16808576583862306
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,1,2,balanced,4.905573120117188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,1,2,balanced,0.30598016738891604
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,1,2,balanced,0.3083097648620606
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,1,2,balanced,0.311910400390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,1,2,balanced,0.31455871582031253
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,1,2,balanced,0.3432486343383789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,1,2,balanced,0.3437388610839844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,1,2,balanced,0.3473318481445312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,1,2,balanced,0.3499417495727539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,1,2,balanced,0.35325313568115235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,1,2,balanced,0.3557939147949219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,1,2,balanced,0.35990272521972655
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,1,2,balanced,0.36204673767089846
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,1,2,balanced,7.583490600585938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,1,2,balanced,0.37460479736328123
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,1,2,balanced,0.39581569671630856
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,1,2,balanced,0.45368064880371095
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,1,2,balanced,0.47217792510986334
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,1,2,balanced,0.5331520080566406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,1,2,balanced,0.8066957092285156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,1,2,balanced,0.872848663330078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,1,2,balanced,1.271580123901367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,2,1,power_law_1.2,1.3029721069335938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,1,2,balanced,1.626982421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,1,2,balanced,10.17164306640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,1,2,power_law_1.01,0.087903995513916
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,1,2,power_law_1.01,0.12317567825317384
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,1,2,power_law_1.01,0.22590463638305663
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,1,2,power_law_1.01,0.2946892738342285
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,1,2,power_law_1.01,0.36346366882324216
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,1,2,balanced,2.4518336486816406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,1,2,power_law_1.01,0.5198425674438477
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,1,2,power_law_1.01,0.534065933227539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,1,2,power_law_1.01,0.5586739349365234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,1,2,power_law_1.01,0.8177049255371094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,1,2,power_law_1.01,0.8109120178222657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,1,2,power_law_1.01,0.839520034790039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,1,2,power_law_1.01,0.8660646057128906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,1,2,power_law_1.01,0.889148178100586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,1,2,power_law_1.01,0.906640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,1,2,power_law_1.01,0.977923812866211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,1,2,balanced,3.1683840942382813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,1,2,power_law_1.01,0.9987276458740235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,1,2,power_law_1.01,1.0322393798828124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,1,2,power_law_1.01,1.1130035400390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,1,2,power_law_1.01,1.231467514038086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,2,1,power_law_1.2,0.9646982574462891
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,1,2,power_law_1.01,1.5088371276855468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,1,2,power_law_1.01,1.8210560607910158
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,1,2,power_law_1.01,2.468721923828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,1,2,power_law_1.01,0.060241918563842776
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,1,2,power_law_1.01,0.08256640434265136
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,1,2,power_law_1.01,3.0908901977539065
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,1,2,power_law_1.01,0.1180799961090088
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,1,2,power_law_1.01,0.1582425594329834
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,1,2,power_law_1.01,0.17897216796874998
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,1,2,power_law_1.01,0.2664255905151367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,1,2,power_law_1.01,4.510616455078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,1,2,power_law_1.01,0.2802380752563477
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,1,2,power_law_1.01,0.29795711517333984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,1,2,power_law_1.01,0.3262656021118164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,1,2,power_law_1.01,0.3374016189575195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,1,2,power_law_1.01,5.849139404296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,1,2,power_law_1.01,0.344716796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,1,2,power_law_1.01,0.34787841796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,1,2,power_law_1.01,0.3507289505004883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,1,2,power_law_1.01,0.35496704101562504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,1,2,power_law_1.01,0.36109569549560544
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,1,2,power_law_1.01,0.38539134979248046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,1,2,power_law_1.01,0.4097139358520508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,1,2,power_law_1.01,0.463641586303711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,1,2,power_law_1.01,0.5428224182128906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,1,2,power_law_1.01,0.6165094375610352
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,1,2,power_law_1.01,8.468577270507812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,1,2,power_law_1.01,0.8505433654785156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,1,2,power_law_1.01,1.016552963256836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,1,2,power_law_1.01,1.5130482482910157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,1,2,power_law_1.2,0.08853504180908203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,1,2,power_law_1.2,0.1119654369354248
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,1,2,power_law_1.01,2.2547789001464844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,1,2,power_law_1.2,0.21856256484985354
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,1,2,power_law_1.2,0.26549631118774414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,1,2,power_law_1.01,11.516782226562501
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,1,2,power_law_1.01,3.0944857788085938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,1,2,power_law_1.2,0.334202880859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,1,2,power_law_1.2,0.5183168029785156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,1,2,power_law_1.2,0.548977928161621
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,1,2,power_law_1.2,0.5793190383911133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,1,2,power_law_1.2,0.8223551940917968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,1,2,power_law_1.01,3.688198547363281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,1,2,power_law_1.2,0.8304204559326172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,1,2,power_law_1.2,0.8599782562255859
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,1,2,power_law_1.2,0.8714304351806641
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,1,2,power_law_1.2,0.8724928283691407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,1,2,power_law_1.2,0.991756820678711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,1,2,power_law_1.2,0.9824626922607422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,1,2,power_law_1.2,1.013839340209961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,1,2,power_law_1.2,1.1501875305175782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,1,2,power_law_1.2,0.9631027221679688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,1,2,power_law_1.2,1.2566912078857422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,1,2,power_law_1.2,1.5861824035644532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,2,1,power_law_1.2,1.703253173828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,1,2,power_law_1.2,1.9208781433105468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,1,2,power_law_1.01,0.33075199127197263
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,1,2,power_law_1.2,0.05608575820922852
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,1,2,power_law_1.2,0.0747929620742798
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,1,2,power_law_1.2,0.10863871574401855
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,1,2,power_law_1.2,2.5375769042968748
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,1,2,power_law_1.2,0.13310848236083986
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,1,2,power_law_1.2,3.1600384521484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,1,2,power_law_1.2,0.18136192321777345
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,1,2,power_law_1.2,0.265739517211914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,1,2,power_law_1.2,0.28924671173095706
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,1,2,power_law_1.2,0.2971379280090332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,1,2,power_law_1.2,0.3328550338745117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,1,2,power_law_1.2,0.3364787292480469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,1,2,power_law_1.2,0.3420300674438477
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,1,2,power_law_1.2,0.3446438217163086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,1,2,power_law_1.2,0.34740734100341797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,1,2,power_law_1.2,0.35747329711914066
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,1,2,power_law_1.2,0.3661209487915039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,1,2,power_law_1.2,0.3764646530151367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,1,2,power_law_1.2,0.3877926254272461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,1,2,power_law_1.2,0.4231078338623047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,1,2,power_law_1.2,0.46784385681152346
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,1,2,power_law_1.2,0.5638515090942382
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,1,2,power_law_1.2,0.638782730102539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,1,2,power_law_1.2,0.861502685546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,2,1,power_law_1.2,1.2487731170654297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,1,2,power_law_1.2,1.1361740875244142
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,1,2,power_law_1.2,1.6015565490722659
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,1,2,power_law_1.2,5.844078369140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,2,1,balanced,0.10409407615661621
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,2,1,balanced,0.16563072204589843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,1,2,power_law_1.2,4.738269348144532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,2,1,balanced,0.05920127868652344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,1,2,power_law_1.2,1.8861964416503905
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,2,1,balanced,0.5732486343383789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,2,1,balanced,0.5798207855224609
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,2,1,balanced,0.5873651123046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,1,2,power_law_1.2,2.8063909912109373
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,2,1,balanced,0.5824716949462891
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,2,1,balanced,0.5899219131469726
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,2,1,balanced,0.5876588821411133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,2,1,balanced,0.3084774398803711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,2,1,balanced,0.5927769470214843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,2,1,balanced,0.6281619262695313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,2,1,balanced,0.6315897750854492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,2,1,balanced,0.6436390686035156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,2,1,balanced,0.5705548858642577
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,2,1,balanced,0.5848704147338867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,2,1,balanced,0.7608620452880859
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,2,1,balanced,0.6618950653076172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,2,1,balanced,0.8228518676757812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,2,1,balanced,1.2634111785888673
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,2,1,balanced,1.3427949523925782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,2,1,balanced,2.0014093017578127
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,2,1,balanced,2.523337554931641
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,2,1,balanced,0.04921599864959717
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,1,2,power_law_1.2,8.628229370117188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,2,1,balanced,0.06753215789794922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,2,1,balanced,0.11008831977844238
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,2,1,balanced,0.1788979148864746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,2,1,balanced,0.3179007911682129
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,2,1,balanced,0.31917951583862303
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,2,1,balanced,0.3255251312255859
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,2,1,balanced,0.32833343505859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,2,1,balanced,0.33129856109619144
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,2,1,balanced,3.790361633300781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,2,1,balanced,0.33311550140380863
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,2,1,balanced,0.3352934265136719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,2,1,balanced,0.6207513427734375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,2,1,balanced,0.3771609497070313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,2,1,balanced,0.3737510299682617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,2,1,balanced,0.377564811706543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,2,1,balanced,0.3804972839355469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,2,1,balanced,0.38547904968261715
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,2,1,balanced,0.3920518493652344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,2,1,balanced,0.40976833343505864
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,2,1,balanced,0.7327724456787109
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,2,1,balanced,0.4280505752563476
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,2,1,balanced,0.48904064178466794
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,2,1,balanced,0.5128665542602538
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,2,1,balanced,0.7757817840576171
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,2,1,balanced,4.901990966796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,2,1,balanced,0.8857913970947265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,2,1,power_law_1.01,0.0600486421585083
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,2,1,power_law_1.01,0.10280256271362305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,2,1,power_law_1.01,0.2416671943664551
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,2,1,balanced,1.7594476318359376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,2,1,balanced,1.2690668487548828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,2,1,balanced,2.397345275878906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,2,1,power_law_1.01,0.4869510269165039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,2,1,power_law_1.01,0.5177337646484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,2,1,power_law_1.01,0.5331852722167969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,2,1,power_law_1.01,0.555978889465332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,2,1,power_law_1.01,0.5478150558471679
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,2,1,balanced,3.170859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,2,1,power_law_1.01,0.5850265502929688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,2,1,power_law_1.01,0.599051513671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,2,1,power_law_1.01,0.6172844696044921
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,2,1,power_law_1.01,0.6689004516601562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,2,1,power_law_1.01,0.1622617530822754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,2,1,power_law_1.01,0.7127769470214844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,2,1,power_law_1.01,0.6573490905761719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,2,1,power_law_1.01,0.7159174346923829
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,2,1,power_law_1.01,0.7716460418701172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,1,2,power_law_1.2,3.8519961547851564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,2,1,power_law_1.01,0.9450387573242187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,2,1,power_law_1.01,1.0911135864257813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,2,1,power_law_1.01,0.31457408905029294
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,2,1,power_law_1.01,1.3637965393066405
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,2,1,power_law_1.01,0.04846848011016845
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,2,1,power_law_1.01,1.640247039794922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,2,1,power_law_1.01,0.06723519802093506
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,2,1,power_law_1.01,0.10960576057434082
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,2,1,power_law_1.01,0.15045503616333006
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,2,1,power_law_1.01,2.294508819580078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,2,1,power_law_1.01,0.1836191940307617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,2,1,power_law_1.01,0.27006399154663085
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,2,1,power_law_1.01,0.29562496185302733
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,2,1,power_law_1.01,0.30027328491210936
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,2,1,power_law_1.01,0.310392951965332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,2,1,power_law_1.01,0.3108064079284668
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,2,1,power_law_1.01,0.32062591552734376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,2,1,power_law_1.01,0.35847488403320316
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,2,1,power_law_1.01,2.9387091064453124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,2,1,power_law_1.01,0.36975807189941406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,2,1,power_law_1.01,0.3801375961303711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,2,1,power_law_1.01,0.38524608612060546
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,2,1,power_law_1.01,4.120768127441407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,2,1,power_law_1.01,0.403699836730957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,2,1,power_law_1.01,0.36170047760009766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,2,1,power_law_1.01,0.4254860687255859
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,2,1,power_law_1.01,0.46993278503417973
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,2,1,power_law_1.01,0.5426956939697265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,1,2,power_law_1.2,12.842847900390627
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,2,1,power_law_1.01,0.6318739318847657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,2,1,power_law_1.01,0.8675027465820312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,2,1,power_law_1.01,5.328983764648437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,2,1,power_law_1.01,1.0228844451904298
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,2,1,power_law_1.2,0.06234432220458984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,2,1,power_law_1.2,0.10202495574951172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,2,1,power_law_1.01,1.4369209289550782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,2,1,power_law_1.2,0.29180288314819336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,2,1,power_law_1.01,0.6462067413330078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,2,1,power_law_1.01,1.9249708557128904
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,2,1,power_law_1.2,0.49369087219238283
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,2,1,power_law_1.2,0.523583984375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,2,1,power_law_1.2,0.5426047897338867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,2,1,power_law_1.2,0.5603385543823242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,2,1,power_law_1.2,0.5587923049926757
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,2,1,power_law_1.01,2.6961505126953123
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,2,1,power_law_1.2,0.5856345748901367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,2,1,power_law_1.2,0.6067763137817382
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,2,1,power_law_1.2,0.6286259078979493
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,2,1,power_law_1.01,3.4123239135742187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,2,1,power_law_1.2,0.6500511932373046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,2,1,power_law_1.2,0.7174931335449218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,2,1,power_law_1.2,0.6653311920166016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,2,1,power_law_1.2,0.8282899475097656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,2,1,power_law_1.2,0.1565503978729248
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,2,1,power_law_1.2,0.22135871887207031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,2,1,power_law_1.2,0.953903350830078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,2,1,power_law_1.2,1.1242111968994142
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,2,1,power_law_1.2,1.3666975402832031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,2,1,power_law_1.2,0.04905087947845459
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,2,1,power_law_1.2,0.06741504192352295
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,2,1,power_law_1.2,1.7060928344726562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,2,1,power_law_1.2,0.10527935981750489
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,2,1,power_law_1.2,0.13970239639282228
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,2,1,power_law_1.2,0.16926656723022462
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,2,1,power_law_1.2,0.27364736557006836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,2,1,power_law_1.2,0.29545919418334965
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,2,1,power_law_1.2,0.3057478332519531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,2,1,power_law_1.2,0.31341951370239257
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,2,1,power_law_1.2,0.7281581115722656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,2,1,power_law_1.2,0.3135161590576172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,2,1,power_law_1.2,0.32031425476074216
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,2,1,power_law_1.2,0.37499969482421874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,2,1,power_law_1.2,0.3652582550048828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,2,1,power_law_1.2,0.37455230712890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,2,1,power_law_1.2,0.38874431610107424
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,2,1,power_law_1.2,0.39130687713623047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,2,1,power_law_1.2,2.2976792907714847
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,2,1,power_law_1.2,0.40913665771484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,2,1,power_law_1.2,0.4329644775390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,2,1,power_law_1.2,0.47278465270996095
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,2,1,power_law_1.2,0.5606067276000977
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,2,1,power_law_1.2,0.6461593627929687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,2,1,power_law_1.2,0.674229736328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,2,1,power_law_1.2,0.8477222442626953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,2,1,power_law_1.2,1.0306470489501953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,1,2,balanced,0.12105983734130858
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,1,2,balanced,0.19454208374023438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,1,2,balanced,0.3846527862548828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,2,1,power_law_1.2,2.9046643066406252
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,1,2,balanced,0.7432038116455078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,1,2,balanced,1.4527885437011718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,2,1,power_law_1.2,1.4171795654296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,1,2,balanced,1.7713279724121094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,1,2,balanced,1.7804518127441404
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,1,2,balanced,1.784880676269531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,1,2,balanced,1.7897010803222657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,1,2,balanced,2.8005169677734374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,1,2,balanced,2.829858703613281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,1,2,balanced,2.8418737792968747
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,2,1,power_law_1.2,1.859998016357422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,1,2,balanced,2.888248291015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,1,2,balanced,2.9456500244140624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,1,2,balanced,2.9835903930664065
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,1,2,balanced,3.060119018554688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,1,2,balanced,3.1703283691406248
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,1,2,balanced,3.317593688964844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,1,2,balanced,3.4292953491210936
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,1,2,balanced,3.8721011352539065
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,1,2,balanced,4.057692260742188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,2,1,power_law_1.2,4.129532775878906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,1,2,balanced,5.7363214111328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,1,2,balanced,7.285142822265625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,1,2,balanced,0.07670400142669678
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,1,2,balanced,0.13519231796264647
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,1,2,balanced,0.22756607055664063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,1,2,balanced,0.3849715042114258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,1,2,balanced,0.7238758087158204
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,1,2,balanced,9.934068603515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,2,1,power_law_1.2,2.630370483398438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,1,2,balanced,0.9002674865722657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,1,2,balanced,0.8985330963134766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,1,2,balanced,0.902339859008789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,1,2,balanced,0.9067967987060547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,1,2,balanced,0.9785510253906249
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,1,2,balanced,0.9902489471435547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,1,2,balanced,1.0059967803955079
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,1,2,balanced,0.9995852661132811
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,1,2,balanced,1.0161203002929686
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,1,2,balanced,1.0352652740478514
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,1,2,balanced,1.0249459075927736
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,1,2,balanced,13.77601318359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,1,2,balanced,1.057738265991211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,1,2,balanced,1.1065702056884765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,1,2,balanced,1.1455142211914062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,1,2,balanced,1.2768831634521485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,1,2,balanced,1.333401641845703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,1,2,balanced,2.1419097900390627
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,1,2,balanced,2.2498214721679686
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,1,2,balanced,3.324313659667969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,1,2,power_law_1.01,0.19837823867797852
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,1,2,power_law_1.01,0.37475841522216796
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,1,2,balanced,4.20142578125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,1,2,power_law_1.01,0.524400634765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,1,2,power_law_1.01,0.8030847930908204
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,1,2,power_law_1.01,0.9395289611816405
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,1,2,balanced,20.9256787109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,1,2,balanced,5.587367553710938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,1,2,power_law_1.01,1.4597311401367188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,1,2,power_law_1.01,1.52621826171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,1,2,power_law_1.01,1.5747354125976563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,1,2,balanced,7.578297729492188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,1,2,power_law_1.01,1.6168858337402345
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,2,1,power_law_1.2,5.272467041015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,1,2,power_law_1.01,2.532481231689453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,1,2,power_law_1.01,2.623420104980469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,1,2,power_law_1.01,2.700528564453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,1,2,power_law_1.01,2.725140380859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,1,2,power_law_1.01,2.86218994140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,1,2,balanced,27.521987304687503
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,1,2,power_law_1.01,2.9242572021484374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,1,2,power_law_1.01,3.063975830078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,1,2,power_law_1.01,3.177076416015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,1,2,power_law_1.01,3.4200076293945316
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,1,2,power_law_1.01,3.6265728759765623
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,1,2,power_law_1.01,4.203719787597656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,1,2,power_law_1.01,4.942452392578125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,1,2,power_law_1.01,0.1364863967895508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,2,1,power_law_1.2,3.381922607421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,1,2,power_law_1.01,0.22025215148925784
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,1,2,power_law_1.01,6.414402465820312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,1,2,power_law_1.01,0.3075494384765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,1,2,power_law_1.01,0.4116723251342774
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,1,2,power_law_1.01,0.4814822387695313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,1,2,power_law_1.01,0.7716799926757812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,1,2,power_law_1.01,8.010682983398437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,1,2,power_law_1.01,0.8154303741455078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,1,2,power_law_1.01,0.8184588623046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,1,2,power_law_1.01,0.8175782775878906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,1,2,power_law_1.01,0.8757683563232422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,1,2,power_law_1.01,0.9259430694580079
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,1,2,power_law_1.01,0.9420902252197265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,1,2,power_law_1.01,0.9155494689941406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,1,2,power_law_1.01,0.9617919921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,1,2,power_law_1.01,0.9734028625488282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,1,2,power_law_1.01,1.0200268554687502
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,1,2,power_law_1.01,1.0282252502441405
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,1,2,power_law_1.01,1.0680831909179687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,1,2,power_law_1.01,1.1811353302001952
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,1,2,power_law_1.01,11.7640673828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,1,2,power_law_1.01,1.4031884765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,1,2,power_law_1.01,1.6794650268554687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,1,2,power_law_1.01,2.124853820800781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,1,2,power_law_1.01,2.455512390136719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,1,2,power_law_1.01,13.533319091796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,1,2,power_law_1.2,0.1949465560913086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,1,2,power_law_1.2,0.35008255004882816
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,1,2,power_law_1.01,3.596641235351562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,1,2,power_law_1.2,0.5224217605590821
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,1,2,power_law_1.2,0.7074598693847657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,1,2,power_law_1.01,4.427082214355469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,1,2,power_law_1.2,0.9770342254638672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,1,2,power_law_1.2,1.4376499938964844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,1,2,power_law_1.2,1.5092108154296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,1,2,power_law_1.2,1.6051776123046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,1,2,power_law_1.2,1.6378892517089845
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,1,2,power_law_1.01,6.591364135742188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,1,2,power_law_1.2,2.551727294921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,1,2,power_law_1.2,2.68245361328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,1,2,power_law_1.2,2.726666259765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,1,2,power_law_1.2,2.818443603515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,1,2,power_law_1.01,20.502302246093752
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,1,2,power_law_1.01,8.561270751953126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,1,2,power_law_1.2,2.8999282836914064
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,1,2,power_law_1.2,2.986003112792969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,1,2,power_law_1.2,3.1245401000976565
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,1,2,power_law_1.2,3.2625076293945314
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,1,2,power_law_1.2,3.503525085449219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,1,2,power_law_1.2,3.785264587402344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,1,2,power_law_1.2,4.443061828613281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,1,2,power_law_1.2,5.0888845825195315
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,1,2,power_law_1.2,0.13634688377380372
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,1,2,power_law_1.2,0.17404159545898437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,1,2,power_law_1.01,27.303449707031252
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,1,2,power_law_1.2,0.28710784912109377
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,1,2,power_law_1.2,6.6174121093750005
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,1,2,power_law_1.2,0.38995201110839844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,1,2,power_law_1.2,0.49732608795166017
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,1,2,power_law_1.2,0.7312127685546874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,1,2,power_law_1.2,0.821923828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,1,2,power_law_1.2,8.222189941406251
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,1,2,power_law_1.2,0.8687423706054688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,1,2,power_law_1.2,0.921102066040039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,1,2,power_law_1.2,0.938387222290039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,1,2,power_law_1.2,0.9587865447998046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,1,2,power_law_1.2,0.963904037475586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,1,2,power_law_1.2,0.9804876708984376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,1,2,power_law_1.2,1.0069913482666015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,1,2,power_law_1.2,0.9839820861816406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,1,2,power_law_1.2,1.065265884399414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,1,2,power_law_1.2,1.1578880310058595
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,1,2,power_law_1.2,12.21612548828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,1,2,power_law_1.2,1.1766118621826172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,1,2,power_law_1.2,1.4416908264160155
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,1,2,power_law_1.2,13.461495361328124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,1,2,power_law_1.2,0.781317138671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,1,2,power_law_1.2,1.750663604736328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,1,2,power_law_1.2,2.110612487792969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,1,2,power_law_1.2,2.598392333984375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,2,1,balanced,0.12343104362487793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,2,1,balanced,0.20434688568115233
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,2,1,balanced,0.3928966522216797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,1,2,power_law_1.2,3.4490087890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,1,2,power_law_1.2,4.349208374023438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,2,1,balanced,0.7492249298095703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,2,1,balanced,1.456033935546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,2,1,balanced,1.7763775634765626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,2,1,balanced,1.7881376647949216
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,2,1,balanced,1.805113525390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,2,1,balanced,1.7928767395019531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,2,1,balanced,1.8133747863769532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,2,1,balanced,1.827751007080078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,2,1,balanced,1.863558349609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,1,2,power_law_1.2,21.06548828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,2,1,balanced,2.8100180053710937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,1,2,power_law_1.2,6.482722778320313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,2,1,balanced,2.8566131591796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,2,1,balanced,2.907894287109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,2,1,balanced,2.9884524536132813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,1,2,power_law_1.2,8.348976440429688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,2,1,balanced,3.109618530273438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,2,1,balanced,3.280048522949219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,2,1,balanced,3.407687072753906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,2,1,balanced,4.094714965820312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,2,1,balanced,4.084617004394532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,2,1,balanced,0.0878758430480957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,2,1,balanced,0.146746883392334
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,2,1,balanced,0.2300876808166504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,2,1,balanced,5.816235961914062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,2,1,balanced,0.39136447906494143
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,2,1,balanced,0.7395180511474609
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,2,1,balanced,0.9122156524658204
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,2,1,balanced,0.9201216125488282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,2,1,balanced,0.9202067565917968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,1,2,power_law_1.2,30.24775390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,2,1,balanced,7.165028686523438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,2,1,balanced,0.927391357421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,2,1,balanced,0.9300128173828124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,2,1,balanced,0.9402317047119141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,2,1,balanced,0.9458035278320314
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,2,1,balanced,1.0469728088378907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,2,1,balanced,1.0567295837402344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,2,1,balanced,1.0753453063964844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,2,1,balanced,1.0714086151123046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,2,1,balanced,9.946039428710938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,2,1,balanced,1.1060684967041017
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,2,1,balanced,1.128091506958008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,2,1,balanced,1.1771110534667968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,2,1,balanced,1.229386215209961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,2,1,balanced,1.3986335754394532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,2,1,balanced,2.120818634033203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,2,1,balanced,2.5139340209960936
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,2,1,balanced,13.055616455078127
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,2,1,power_law_1.01,0.12966783523559572
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,2,1,power_law_1.01,0.20298431396484373
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,2,1,power_law_1.01,0.39451904296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,2,1,balanced,3.405518798828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,2,1,power_law_1.01,0.6314271926879883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,2,1,power_law_1.01,0.8578521728515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,2,1,balanced,4.292035217285156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,2,1,power_law_1.01,1.3999807739257812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,2,1,power_law_1.01,1.487889862060547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,2,1,power_law_1.01,1.540771789550781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,2,1,power_law_1.01,1.5925350952148438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,2,1,balanced,5.690851440429688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,2,1,power_law_1.01,1.6365895080566406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,2,1,power_law_1.01,1.7047769165039064
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,2,1,power_law_1.01,1.8158277893066406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,2,1,balanced,18.566990966796872
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,2,1,power_law_1.01,2.6411090087890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,2,1,balanced,7.691153564453126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,2,1,power_law_1.01,2.773145751953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,2,1,power_law_1.01,2.842706604003906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,2,1,power_law_1.01,2.876629638671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,2,1,power_law_1.01,3.031715087890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,2,1,power_law_1.01,3.3555328369140627
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,2,1,power_law_1.01,3.5316967773437495
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,2,1,power_law_1.01,4.01686279296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,2,1,power_law_1.01,4.625537414550782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,2,1,power_law_1.01,0.08649087905883789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,2,1,power_law_1.01,0.14516863822937012
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,2,1,balanced,25.025810546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,2,1,power_law_1.01,0.22946048736572267
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,2,1,power_law_1.01,5.895598754882813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,2,1,power_law_1.01,0.35213504791259764
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,2,1,power_law_1.01,0.4507839965820312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,2,1,power_law_1.01,0.7194745635986328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,2,1,power_law_1.01,0.7632755279541016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,2,1,power_law_1.01,0.7932991790771484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,2,1,power_law_1.01,0.8185523223876953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,2,1,power_law_1.01,0.8330016326904296
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,2,1,power_law_1.01,7.194520874023437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,2,1,power_law_1.01,0.8541855621337892
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,2,1,power_law_1.01,0.8741779327392578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,2,1,power_law_1.01,0.9393286132812501
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,2,1,power_law_1.01,0.9872550201416015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,2,1,power_law_1.01,1.0024460601806642
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,2,1,power_law_1.01,1.0141664123535157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,2,1,power_law_1.01,1.0615213012695313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,2,1,power_law_1.01,10.011392211914062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,2,1,power_law_1.01,1.1294239807128907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,2,1,power_law_1.01,1.2631180572509766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,2,1,power_law_1.01,1.3886444091796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,2,1,power_law_1.01,1.6672154235839844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,2,1,power_law_1.01,1.9967079162597656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,2,1,power_law_1.01,2.457861175537109
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,2,1,power_law_1.01,12.663948974609374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,2,1,power_law_1.2,0.12094783782958984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,2,1,power_law_1.2,0.20383424758911134
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,2,1,power_law_1.2,0.3886854553222656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,2,1,power_law_1.2,0.5827020645141602
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,2,1,power_law_1.01,3.4353842163085937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,2,1,power_law_1.2,0.8047110748291015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,2,1,power_law_1.01,4.119084167480469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,2,1,power_law_1.2,1.3696102905273437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,2,1,power_law_1.2,1.4865811157226563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,2,1,power_law_1.2,1.5655955505371093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,2,1,power_law_1.2,1.6314874267578126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,2,1,power_law_1.01,5.860169677734374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,2,1,power_law_1.2,1.6806022644042968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,2,1,power_law_1.2,1.7781759643554687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,2,1,power_law_1.2,1.8664378356933593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,2,1,power_law_1.01,18.27994384765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,2,1,power_law_1.01,7.724851684570313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,2,1,power_law_1.2,2.732815856933594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,2,1,power_law_1.2,2.8041894531250002
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,2,1,power_law_1.2,2.9501605224609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,2,1,power_law_1.2,3.0075558471679686
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,2,1,power_law_1.2,3.1326190185546876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,2,1,power_law_1.2,3.349760131835937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,2,1,power_law_1.2,3.5351821899414064
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,2,1,power_law_1.2,4.124366149902344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,2,1,power_law_1.01,23.941462402343753
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,2,1,power_law_1.2,4.685650024414063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,2,1,power_law_1.2,0.08775168418884277
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,2,1,power_law_1.2,0.14613183975219726
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,2,1,power_law_1.2,0.22628799438476563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,2,1,power_law_1.2,0.3104870414733887
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,2,1,power_law_1.2,6.007698364257813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,2,1,power_law_1.2,0.4190291213989258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,2,1,power_law_1.2,0.7091814422607422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,2,1,power_law_1.2,0.7551366424560546
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,2,1,power_law_1.2,0.8031378936767577
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,2,1,power_law_1.2,7.415847167968749
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,2,1,power_law_1.2,0.8280307006835936
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,2,1,power_law_1.2,0.8483462524414062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,2,1,power_law_1.2,0.8797574615478515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,2,1,power_law_1.2,0.8913107299804688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,2,1,power_law_1.2,0.9994828796386719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,2,1,power_law_1.2,1.0120832061767577
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,2,1,power_law_1.2,1.045604476928711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,2,1,power_law_1.2,1.0717932891845703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,2,1,power_law_1.2,1.0876531219482422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,2,1,power_law_1.2,1.1637344360351562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,2,1,power_law_1.2,10.204340209960938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,2,1,power_law_1.2,1.2756018829345703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,2,1,power_law_1.2,1.4084576416015626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,2,1,power_law_1.2,1.6787149047851564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,2,1,power_law_1.2,2.067008056640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,2,1,power_law_1.2,2.534643859863281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,1,2,balanced,0.12279040336608886
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,1,2,balanced,0.20682111740112305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,2,1,power_law_1.2,13.107532958984375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,2,1,power_law_1.2,3.4275653076171877
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,1,2,balanced,0.42197246551513673
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,1,2,balanced,0.7036172485351562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,1,2,balanced,1.3354226684570312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,2,1,power_law_1.2,4.240196533203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,1,2,balanced,2.6460696411132814
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,1,2,balanced,3.917369689941406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,1,2,balanced,3.982700805664062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,1,2,balanced,3.9531341552734376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,1,2,balanced,3.976629638671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,1,2,balanced,4.022857055664063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,1,2,balanced,4.024001159667969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,2,1,power_law_1.2,5.906160278320312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,1,2,balanced,4.055155334472657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,2,1,power_law_1.2,18.42805419921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,1,2,balanced,6.196408081054687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,1,2,balanced,6.19362060546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,1,2,balanced,6.230769653320313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,1,2,balanced,6.376039428710937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,1,2,balanced,6.523707885742188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,1,2,balanced,6.764837036132812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,2,1,power_law_1.2,7.462277221679687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,1,2,balanced,7.1883697509765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,1,2,balanced,7.554916992187501
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,1,2,balanced,0.07085951805114746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,1,2,balanced,8.045145874023437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,1,2,balanced,0.11795455932617187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,1,2,balanced,0.2133452796936035
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,1,2,balanced,0.3945292663574219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,1,2,balanced,9.235197143554688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,1,2,balanced,0.7311743927001954
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,1,2,balanced,1.3183027648925782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,1,2,balanced,9.972501831054688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,1,2,balanced,1.972515869140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,1,2,balanced,1.9673817443847657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,1,2,balanced,1.9661439514160157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,2,1,power_law_1.2,24.1590966796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,1,2,balanced,1.970635528564453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,1,2,balanced,1.9829991149902344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,1,2,balanced,13.733935546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,1,2,balanced,1.9982540893554688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,1,2,balanced,2.1739698791503903
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,1,2,balanced,1.9891506958007814
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,1,2,balanced,2.1933581542968748
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,1,2,balanced,2.2074163818359374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,1,2,balanced,2.2099455261230467
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,1,2,balanced,2.3826905822753908
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,1,2,balanced,2.374266815185547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,1,2,balanced,2.4114393615722656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,1,2,balanced,2.6450701904296876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,1,2,balanced,2.7734130859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,1,2,balanced,2.936645202636719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,1,2,power_law_1.01,0.2190540885925293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,1,2,balanced,19.194072265625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,1,2,power_law_1.01,0.43021183013916015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,1,2,balanced,3.590719909667969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,1,2,power_law_1.01,0.6906124877929687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,1,2,power_law_1.01,0.914751968383789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,1,2,balanced,5.107234497070313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,1,2,power_law_1.01,1.5262220764160157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,1,2,power_law_1.01,1.9992767333984376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,1,2,balanced,5.599566040039063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,1,2,power_law_1.01,2.2181939697265625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,1,2,balanced,7.64745361328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,1,2,power_law_1.01,2.973546142578125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,1,2,power_law_1.01,3.014964599609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,1,2,power_law_1.01,3.0475622558593747
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,1,2,power_law_1.01,3.3243481445312497
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,1,2,power_law_1.01,2.82922119140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,1,2,balanced,27.640458984374998
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,1,2,power_law_1.01,3.394415283203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,1,2,power_law_1.01,5.186165771484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,1,2,power_law_1.01,5.180582275390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,1,2,power_law_1.01,5.317199096679688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,1,2,power_law_1.01,5.427498168945313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,1,2,power_law_1.01,6.005987548828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,1,2,power_law_1.01,6.145861206054688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,1,2,power_law_1.01,6.901153564453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,1,2,power_law_1.01,0.11822336196899415
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,1,2,power_law_1.01,0.21230848312377928
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,1,2,power_law_1.01,7.525488891601563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,1,2,power_law_1.01,0.3929536056518555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,1,2,power_law_1.01,0.5257535934448242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,1,2,power_law_1.01,8.684732055664062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,1,2,power_law_1.01,0.7972684478759765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,1,2,power_law_1.01,9.918274536132813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,1,2,power_law_1.01,1.019384307861328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,1,2,power_law_1.01,1.1347776031494141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,1,2,power_law_1.01,1.4351155090332033
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,1,2,power_law_1.01,1.5023974609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,1,2,power_law_1.01,1.5316569519042968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,1,2,power_law_1.01,1.5397747802734376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,1,2,power_law_1.01,12.65877685546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,1,2,power_law_1.01,1.629669189453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,1,2,power_law_1.01,1.6441766357421876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,1,2,power_law_1.01,1.8158872985839842
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,1,2,power_law_1.01,1.8198220825195315
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,1,2,power_law_1.01,1.851393280029297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,1,2,power_law_1.01,1.9173887634277342
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,1,2,power_law_1.01,15.405423583984376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,1,2,power_law_1.01,1.9700172424316407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,1,2,power_law_1.01,2.1553446960449216
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,1,2,power_law_1.01,2.2393650817871094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,1,2,power_law_1.01,2.4198399353027344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,1,2,power_law_1.01,2.9417575073242186
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,1,2,power_law_1.01,3.1827301025390624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,1,2,power_law_1.2,0.21869184494018556
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,1,2,power_law_1.2,0.4281318283081054
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,1,2,power_law_1.2,0.6216064071655273
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,1,2,power_law_1.01,4.221415710449219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,1,2,power_law_1.01,20.458067626953124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,1,2,power_law_1.2,0.9043891143798828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,1,2,power_law_1.2,1.4151206970214845
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,1,2,power_law_1.01,5.234089965820313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,1,2,power_law_1.2,1.7996928405761718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,1,2,power_law_1.2,2.2179046630859376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,1,2,power_law_1.2,2.7746597290039063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,1,2,power_law_1.2,2.8648999023437502
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,1,2,power_law_1.2,3.093367004394531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,1,2,power_law_1.2,3.072340393066406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,1,2,power_law_1.01,8.846539306640626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,1,2,power_law_1.2,3.239851379394531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,1,2,power_law_1.01,26.407260742187496
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,1,2,power_law_1.2,3.3615847778320314
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,1,2,power_law_1.2,5.056514587402344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,1,2,power_law_1.2,5.117018737792969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,1,2,power_law_1.2,5.317205810546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,1,2,power_law_1.2,5.431588134765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,1,2,power_law_1.2,6.062872314453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,1,2,power_law_1.2,6.238115844726563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,1,2,power_law_1.01,6.651932373046876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,1,2,power_law_1.2,7.0236083984375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,1,2,power_law_1.2,0.11799679756164552
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,1,2,power_law_1.2,0.21259263992309568
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,1,2,power_law_1.2,0.32968448638916015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,1,2,power_law_1.2,7.921489868164062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,1,2,power_law_1.2,0.5139929580688476
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,1,2,power_law_1.2,9.212560424804687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,1,2,power_law_1.2,0.7401856231689453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,1,2,power_law_1.2,0.9558322906494141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,1,2,power_law_1.2,10.518046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,1,2,power_law_1.2,1.0614527893066406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,1,2,power_law_1.2,1.431530303955078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,1,2,power_law_1.2,1.4714752197265626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,1,2,power_law_1.2,1.4382949829101563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,1,2,power_law_1.2,1.5998439025878906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,1,2,power_law_1.2,13.101207275390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,1,2,power_law_1.2,1.5837210083007813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,1,2,power_law_1.2,1.637873992919922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,1,2,power_law_1.2,1.785364532470703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,1,2,power_law_1.2,1.8534643554687498
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,1,2,power_law_1.2,1.837767639160156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,1,2,power_law_1.2,1.8806515502929688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,1,2,power_law_1.2,2.0227366638183595
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,1,2,power_law_1.2,2.186090240478516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,1,2,power_law_1.2,16.365687255859378
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,1,2,power_law_1.2,2.3760780334472655
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,1,2,power_law_1.2,2.5282162475585936
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,1,2,power_law_1.2,3.011956481933594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,1,2,power_law_1.2,3.45236083984375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,2,1,balanced,0.14060735702514648
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,2,1,balanced,0.20855039596557617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,2,1,balanced,0.4213075256347656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,1,2,power_law_1.2,4.445633239746094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,2,1,balanced,0.7020460510253906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,1,2,power_law_1.2,5.330245361328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,1,2,power_law_1.2,21.558125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,2,1,balanced,1.3409100341796876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,1,2,power_law_1.2,7.18695068359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,2,1,balanced,2.6851455688476564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,1,2,power_law_1.2,8.927065429687499
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,2,1,balanced,3.973624267578125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,2,1,balanced,3.9677511596679684
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,2,1,balanced,3.9865246582031246
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,2,1,balanced,4.072015380859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,1,2,power_law_1.2,28.26865234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,2,1,balanced,4.051937866210937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,2,1,balanced,4.075797119140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,2,1,balanced,4.011149291992187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,2,1,balanced,4.238976745605468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,2,1,balanced,4.104010314941407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,2,1,balanced,4.185640258789062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,2,1,balanced,6.195653686523437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,2,1,balanced,6.345150756835937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,2,1,balanced,6.577388305664063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,2,1,balanced,7.066077270507813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,2,1,balanced,0.08238207817077636
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,2,1,balanced,0.13096192359924316
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,2,1,balanced,0.2224511909484863
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,2,1,balanced,7.485170288085937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,2,1,balanced,0.40271232604980467
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,2,1,balanced,8.002863159179688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,2,1,balanced,0.7108672332763672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,2,1,balanced,9.242616577148437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,2,1,balanced,1.3443359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,2,1,balanced,9.802346191406249
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,2,1,balanced,2.103246765136719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,2,1,balanced,1.9873394775390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,2,1,balanced,2.042297668457031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,2,1,balanced,2.0157562255859376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,2,1,balanced,2.0331436157226563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,2,1,balanced,2.0461529541015624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,2,1,balanced,2.0283859252929686
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,2,1,balanced,2.053907775878906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,2,1,balanced,2.0778349304199217
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,2,1,balanced,2.085973052978516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,2,1,balanced,2.397696075439453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,2,1,balanced,2.4743174743652343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,2,1,balanced,13.6663916015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,2,1,balanced,2.4770828247070313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,2,1,balanced,2.7026046752929687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,2,1,balanced,2.666375732421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,2,1,balanced,2.9587640380859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,2,1,balanced,3.0105203247070316
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,2,1,power_law_1.01,0.13343680381774903
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,2,1,power_law_1.01,0.20652416229248044
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,2,1,balanced,3.590458984375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,2,1,power_law_1.01,0.41935039520263667
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,2,1,power_law_1.01,0.7016550445556641
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,2,1,balanced,17.855006103515628
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,2,1,power_law_1.01,1.2529843139648438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,2,1,balanced,5.167849731445313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,2,1,power_law_1.01,1.752791748046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,2,1,power_law_1.01,2.023670349121094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,2,1,balanced,5.861063842773437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,2,1,power_law_1.01,2.636138305664063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,2,1,power_law_1.01,2.77626953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,2,1,power_law_1.01,2.8868453979492186
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,2,1,balanced,7.842529296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,2,1,power_law_1.01,3.004432067871094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,2,1,power_law_1.01,3.0306079101562498
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,2,1,power_law_1.01,3.2585931396484376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,2,1,power_law_1.01,3.4485964965820317
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,2,1,power_law_1.01,3.5420263671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,2,1,power_law_1.01,3.7892327880859376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,2,1,balanced,26.184809570312503
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,2,1,power_law_1.01,5.208413696289062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,2,1,power_law_1.01,5.643397216796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,2,1,power_law_1.01,5.985370483398437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,2,1,power_law_1.01,6.726243286132812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,2,1,power_law_1.01,0.08294015884399414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,2,1,power_law_1.01,7.18028076171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,2,1,power_law_1.01,0.1304543972015381
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,2,1,power_law_1.01,0.2235366439819336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,2,1,power_law_1.01,0.40335166931152344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,2,1,power_law_1.01,8.247953491210938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,2,1,power_law_1.01,0.6720108795166015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,2,1,power_law_1.01,0.9079232025146485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,2,1,power_law_1.01,1.0378797149658203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,2,1,power_law_1.01,1.3374610900878907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,2,1,power_law_1.01,9.436573486328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,2,1,power_law_1.01,1.4101068115234374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,2,1,power_law_1.01,1.476324462890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,2,1,power_law_1.01,1.5259014892578127
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,2,1,power_law_1.01,1.544058837890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,2,1,power_law_1.01,1.6062118530273437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,2,1,power_law_1.01,11.721866455078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,2,1,power_law_1.01,1.655292205810547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,2,1,power_law_1.01,1.6723622131347657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,2,1,power_law_1.01,1.7286335754394533
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,2,1,power_law_1.01,1.9865759277343749
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,2,1,power_law_1.01,2.0521145629882813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,2,1,power_law_1.01,2.1914540100097657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,2,1,power_law_1.01,14.306962890624998
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,2,1,power_law_1.01,2.3751315307617187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,2,1,power_law_1.01,2.4848095703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,2,1,power_law_1.01,2.9653408813476565
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,2,1,power_law_1.2,0.14017087936401368
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,2,1,power_law_1.01,3.319671020507813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,2,1,power_law_1.2,0.19580095291137695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,2,1,power_law_1.2,0.4248640060424805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,2,1,power_law_1.01,4.269450378417969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,2,1,power_law_1.2,0.7010131072998047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,2,1,power_law_1.01,5.049432373046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,2,1,power_law_1.2,1.148059539794922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,2,1,power_law_1.01,19.463397216796874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,2,1,power_law_1.2,1.5969850158691405
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,2,1,power_law_1.2,1.8634144592285158
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,2,1,power_law_1.01,6.823325195312501
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,2,1,power_law_1.2,2.5738470458984373
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,2,1,power_law_1.2,2.6833099365234374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,2,1,power_law_1.2,2.7760537719726566
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,2,1,power_law_1.01,8.315975952148438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,2,1,power_law_1.2,2.9024954223632813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,2,1,power_law_1.2,3.04638916015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,2,1,power_law_1.01,24.38577392578125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,2,1,power_law_1.2,3.154963073730469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,2,1,power_law_1.2,3.35458251953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,2,1,power_law_1.2,3.548359069824219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,2,1,power_law_1.2,3.7403256225585935
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,2,1,power_law_1.2,5.253897094726563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,2,1,power_law_1.2,5.76412109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,2,1,power_law_1.2,6.145587158203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,2,1,power_law_1.2,6.724030151367188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,2,1,power_law_1.2,0.08253567695617677
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,2,1,power_law_1.2,7.555993041992187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,2,1,power_law_1.2,0.1310431957244873
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,2,1,power_law_1.2,0.2234873580932617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,2,1,power_law_1.2,0.4037715148925781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,2,1,power_law_1.2,8.571980590820313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,2,1,power_law_1.2,0.6196908950805664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,2,1,power_law_1.2,0.8338400268554688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,2,1,power_law_1.2,0.9652550506591797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,2,1,power_law_1.2,9.692269287109374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,2,1,power_law_1.2,1.3061593627929686
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,2,1,power_law_1.2,1.366903076171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,2,1,power_law_1.2,1.4195397949218749
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,2,1,power_law_1.2,1.4618150329589843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,2,1,power_law_1.2,12.164671630859376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,2,1,power_law_1.2,1.5312063598632812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,2,1,power_law_1.2,1.5704045104980469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,2,1,power_law_1.2,1.6303045654296873
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,2,1,power_law_1.2,1.6628550720214843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,2,1,power_law_1.2,1.7021817016601564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,2,1,power_law_1.2,1.9863250732421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,2,1,power_law_1.2,2.0902803039550784
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,2,1,power_law_1.2,2.2339251708984373
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,2,1,power_law_1.2,2.4209382629394534
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,2,1,power_law_1.2,14.71858642578125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,2,1,power_law_1.2,2.6431533813476564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,2,1,power_law_1.2,3.1404428100585937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,1,2,balanced,0.05972864151000976
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,1,2,balanced,0.08539008140563965
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,1,2,balanced,0.12425472259521483
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,1,2,balanced,0.22768512725830078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,2,1,power_law_1.2,3.4457940673828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,1,2,balanced,0.43584384918212893
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,1,2,balanced,0.857729263305664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,1,2,balanced,0.8540966033935546
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,1,2,balanced,0.8593267059326172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,2,1,power_law_1.2,4.465280151367187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,1,2,balanced,0.861550064086914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,1,2,balanced,0.8658585357666017
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,1,2,balanced,0.8737484741210938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,2,1,power_law_1.2,5.1683197021484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,2,1,power_law_1.2,19.912127685546874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,1,2,balanced,1.3052249145507813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,1,2,balanced,1.3124429321289062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,1,2,balanced,1.327499542236328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,1,2,balanced,1.3764646911621095
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,1,2,balanced,1.3567680358886718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,1,2,balanced,1.4113203430175782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,2,1,power_law_1.2,6.82320556640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,1,2,balanced,1.4926681518554688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,1,2,balanced,1.5533453369140626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,1,2,balanced,1.6535717773437502
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,1,2,balanced,1.7280613708496095
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,1,2,balanced,1.994458923339844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,1,2,balanced,2.173012542724609
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,1,2,balanced,0.04017792224884033
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,1,2,balanced,0.05411839962005616
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,2,1,power_law_1.2,8.524264526367187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,1,2,balanced,0.13985152244567872
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,1,2,balanced,0.24716672897338868
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,1,2,balanced,0.4416998291015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,1,2,balanced,3.10727294921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,1,2,balanced,0.08762495994567872
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,1,2,balanced,0.4477542495727539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,1,2,balanced,0.4477721786499023
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,1,2,balanced,0.4534732818603516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,1,2,balanced,0.45236480712890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,1,2,balanced,0.45614974975585937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,1,2,balanced,3.936588745117187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,1,2,balanced,0.49894783020019534
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,1,2,balanced,0.5004748916625976
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,1,2,balanced,0.5021567916870117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,1,2,balanced,0.5080038452148438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,2,1,power_law_1.2,25.1616357421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,1,2,balanced,0.5125452804565429
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,1,2,balanced,0.5233203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,1,2,balanced,0.5433766555786133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,1,2,balanced,0.6104191970825196
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,1,2,balanced,0.760901107788086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,1,2,balanced,0.7021568298339844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,1,2,balanced,5.755216674804688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,1,2,balanced,0.8547968292236329
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,1,2,power_law_1.01,0.07144576072692871
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,1,2,balanced,0.5842444610595703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,1,2,balanced,1.222840347290039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,1,2,power_law_1.01,0.1255513572692871
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,1,2,power_law_1.01,0.20324352264404294
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,1,2,balanced,1.3897138977050782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,1,2,power_law_1.01,0.3010227203369141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,1,2,power_law_1.01,0.4068147277832031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,1,2,power_law_1.01,0.5096422576904297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,1,2,power_law_1.01,0.6948556518554687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,1,2,power_law_1.01,0.7067558288574218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,1,2,balanced,2.122694396972656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,1,2,power_law_1.01,0.7115993499755859
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,1,2,balanced,7.559668579101563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,1,2,power_law_1.01,0.7168370819091796
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,1,2,power_law_1.01,0.7442201232910156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,1,2,balanced,2.7905191040039066
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,1,2,power_law_1.01,1.111409912109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,1,2,power_law_1.01,1.134566421508789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,1,2,power_law_1.01,1.2064141082763673
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,1,2,power_law_1.01,1.2539852905273439
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,1,2,power_law_1.01,1.2556594848632812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,1,2,power_law_1.01,1.3366963195800783
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,1,2,power_law_1.01,1.3868173217773436
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,1,2,power_law_1.01,1.4823948669433595
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,1,2,power_law_1.01,1.5917004394531251
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,1,2,power_law_1.01,1.7852391052246095
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,1,2,power_law_1.01,0.05208447933197021
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,1,2,power_law_1.01,0.08100607872009277
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,1,2,power_law_1.01,2.203801574707031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,1,2,power_law_1.01,0.11544575691223144
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,1,2,power_law_1.01,0.16359039306640627
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,1,2,power_law_1.01,2.5965618896484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,1,2,power_law_1.01,0.2180019187927246
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,1,2,power_law_1.01,0.25897344589233395
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,1,2,power_law_1.01,0.3626918411254883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,1,2,power_law_1.01,0.3690163040161133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,1,2,power_law_1.01,0.3790719985961914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,1,2,power_law_1.01,0.38257919311523436
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,1,2,power_law_1.01,3.59238525390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,1,2,power_law_1.01,0.395623664855957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,1,2,power_law_1.01,0.4401152038574219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,1,2,power_law_1.01,0.4473715209960938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,1,2,power_law_1.01,0.4515788650512695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,1,2,power_law_1.01,0.4679180908203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,1,2,power_law_1.01,0.46563072204589845
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,1,2,power_law_1.01,0.4876019287109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,1,2,power_law_1.01,0.5124825668334961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,1,2,power_law_1.01,4.528988037109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,1,2,power_law_1.01,0.5513382339477539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,1,2,power_law_1.01,0.6048230361938477
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,1,2,power_law_1.01,0.7022310638427735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,1,2,power_law_1.01,0.8193689727783203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,1,2,power_law_1.01,1.0496575927734375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,1,2,power_law_1.2,0.07273727893829346
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,1,2,power_law_1.2,0.1256486415863037
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,1,2,power_law_1.2,0.2021734428405762
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,1,2,power_law_1.01,6.513502807617188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,1,2,power_law_1.01,1.6449548339843751
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,1,2,power_law_1.2,0.29570688247680665
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,1,2,power_law_1.2,0.38491134643554686
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,1,2,power_law_1.2,0.48276096343994135
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,1,2,power_law_1.01,2.3241293334960935
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,1,2,power_law_1.2,0.7004108428955078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,1,2,power_law_1.2,0.704748764038086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,1,2,power_law_1.2,0.7151270294189453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,1,2,power_law_1.01,3.1050100708007813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,1,2,power_law_1.2,0.7219660949707032
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,1,2,power_law_1.2,0.7705228424072266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,1,2,power_law_1.01,8.425331420898438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,1,2,power_law_1.01,1.3961138916015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,1,2,power_law_1.2,1.1429670715332032
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,1,2,power_law_1.2,1.1664857482910156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,1,2,power_law_1.2,1.241274871826172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,1,2,power_law_1.2,1.2564224243164062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,1,2,power_law_1.2,1.3092083740234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,1,2,power_law_1.2,1.3538111877441406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,1,2,power_law_1.2,1.4507212829589844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,1,2,power_law_1.2,1.514265594482422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,1,2,power_law_1.2,1.671240997314453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,1,2,power_law_1.2,0.052074241638183597
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,1,2,power_law_1.2,1.8734002685546876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,1,2,power_law_1.2,0.08689663887023927
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,1,2,power_law_1.2,0.11768704414367677
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,1,2,power_law_1.2,2.225932769775391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,1,2,power_law_1.2,0.14840831756591796
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,1,2,power_law_1.2,2.718516540527344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,1,2,power_law_1.2,0.1942451286315918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,1,2,power_law_1.2,0.2567731285095215
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,1,2,power_law_1.2,0.3608025741577149
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,1,2,power_law_1.2,0.37126529693603516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,1,2,power_law_1.2,0.37491455078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,1,2,power_law_1.2,0.3860006332397461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,1,2,power_law_1.2,0.44279296874999996
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,1,2,power_law_1.2,0.4487385559082031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,1,2,power_law_1.2,0.4643788909912109
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,1,2,power_law_1.2,0.3784089660644531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,1,2,power_law_1.2,0.4644761657714843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,1,2,power_law_1.2,0.4799398422241211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,1,2,power_law_1.2,4.876802673339844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,1,2,power_law_1.2,3.7539111328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,1,2,power_law_1.2,0.4986240005493164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,1,2,power_law_1.2,0.528856315612793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,1,2,power_law_1.2,0.568985595703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,1,2,power_law_1.2,0.6334143829345703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,1,2,power_law_1.2,0.7029542541503906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,1,2,power_law_1.2,0.8713011169433594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,2,1,balanced,0.05966464042663574
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,1,2,power_law_1.2,1.0502464294433593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,2,1,balanced,0.08085824012756347
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,1,2,power_law_1.2,1.3600627136230468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,1,2,power_law_1.2,6.819879760742188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,2,1,balanced,0.13447936058044435
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,2,1,balanced,0.23805055618286133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,1,2,power_law_1.2,1.7236108398437502
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,2,1,balanced,0.44664768218994144
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,2,1,balanced,0.8620159912109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,2,1,balanced,0.8607481384277345
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,2,1,balanced,0.8649593353271484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,1,2,power_law_1.2,2.56163330078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,2,1,balanced,0.8723097229003907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,2,1,balanced,0.8669446563720703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,2,1,balanced,0.8971327972412109
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,2,1,balanced,0.8868102264404296
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,1,2,power_law_1.2,3.618014831542969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,2,1,balanced,0.8936255645751953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,2,1,balanced,0.9050125122070313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,2,1,balanced,1.3261875915527344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,2,1,balanced,1.3408172607421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,1,2,power_law_1.2,9.131270141601563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,2,1,balanced,1.404388427734375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,2,1,balanced,1.4545971679687502
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,2,1,balanced,1.519471435546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,2,1,balanced,1.6126336669921877
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,2,1,balanced,1.6863693237304687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,2,1,balanced,0.04332672119140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,2,1,balanced,0.06095359802246094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,2,1,balanced,0.09018048286437988
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,2,1,balanced,0.14370112419128417
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,2,1,balanced,0.24887487411499026
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,2,1,balanced,1.9457830810546874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,2,1,balanced,2.1163238525390624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,2,1,balanced,0.45932350158691404
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,2,1,balanced,0.46445697784423834
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,2,1,balanced,0.4707654571533203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,2,1,balanced,0.4715212631225586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,2,1,balanced,0.47229633331298826
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,2,1,balanced,0.47892097473144535
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,2,1,balanced,2.9913510131835936
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,2,1,balanced,0.4829516983032226
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,2,1,balanced,0.48575294494628907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,2,1,balanced,0.4911276626586914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,2,1,balanced,0.5477951812744141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,2,1,balanced,0.5631916809082032
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,2,1,balanced,0.5615763092041016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,2,1,balanced,0.5765651321411133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,2,1,balanced,0.589994888305664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,2,1,balanced,3.7958151245117184
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,2,1,balanced,0.6641836547851563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,2,1,balanced,0.7579808044433594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,2,1,balanced,0.8134207916259765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,2,1,balanced,0.6309465789794922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,2,1,power_law_1.01,0.05592959880828857
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,2,1,balanced,1.2862509155273438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,2,1,balanced,1.4172959899902344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,2,1,power_law_1.01,0.07904895782470703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,2,1,balanced,5.628677978515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,2,1,power_law_1.01,0.13372672080993653
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,2,1,power_law_1.01,0.24060031890869144
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,2,1,power_law_1.01,0.3520735931396485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,2,1,balanced,2.1333772277832033
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,2,1,power_law_1.01,0.4486284637451172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,2,1,power_law_1.01,0.635338249206543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,2,1,power_law_1.01,0.6642848205566406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,2,1,power_law_1.01,0.678503646850586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,2,1,power_law_1.01,0.689715805053711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,2,1,balanced,7.137625732421876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,2,1,power_law_1.01,0.7317177581787109
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,2,1,balanced,2.6499423217773437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,2,1,power_law_1.01,0.7562060546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,2,1,power_law_1.01,0.776780776977539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,2,1,power_law_1.01,0.8172882843017579
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,2,1,power_law_1.01,1.1731616210937499
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,2,1,power_law_1.01,1.194170913696289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,2,1,power_law_1.01,1.2610092926025391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,2,1,power_law_1.01,1.3361196899414063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,2,1,power_law_1.01,1.398937530517578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,2,1,power_law_1.01,1.5264633178710938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,2,1,power_law_1.01,1.670155487060547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,2,1,power_law_1.01,0.04488575935363769
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,2,1,power_law_1.01,0.0623686408996582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,2,1,power_law_1.01,1.9720909118652343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,2,1,power_law_1.01,0.09240320205688476
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,2,1,power_law_1.01,0.14385472297668458
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,2,1,power_law_1.01,0.20646015167236328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,2,1,power_law_1.01,0.2511609649658203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,2,1,power_law_1.01,2.3445216369628907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,2,1,power_law_1.01,0.34713665008544925
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,2,1,power_law_1.01,0.36673793792724607
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,2,1,power_law_1.01,0.3728057479858399
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,2,1,power_law_1.01,0.38177024841308593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,2,1,power_law_1.01,0.3975430297851562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,2,1,power_law_1.01,0.41057086944580073
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,2,1,power_law_1.01,0.41539264678955073
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,2,1,power_law_1.01,0.42280128479003903
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,2,1,power_law_1.01,0.48643840789794923
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,2,1,power_law_1.01,0.48600513458251954
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,2,1,power_law_1.01,3.148822937011719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,2,1,power_law_1.01,0.5079155349731446
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,2,1,power_law_1.01,0.5357900619506836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,2,1,power_law_1.01,0.5681625747680664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,2,1,power_law_1.01,0.6215532684326173
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,2,1,power_law_1.01,0.6992550659179687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,2,1,power_law_1.01,4.103445739746094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,2,1,power_law_1.01,0.8334976196289062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,2,1,power_law_1.01,1.0088902282714844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,2,1,power_law_1.2,0.05763455867767334
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,2,1,power_law_1.01,5.6146173095703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,2,1,power_law_1.2,0.08050111770629884
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,2,1,power_law_1.2,0.13312000274658203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,2,1,power_law_1.01,1.311242218017578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,2,1,power_law_1.2,0.22620735168457032
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,2,1,power_law_1.2,0.3226732635498047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,2,1,power_law_1.01,1.6719993591308593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,2,1,power_law_1.2,0.42442047119140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,2,1,power_law_1.2,0.6279500961303711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,2,1,power_law_1.2,0.6529196929931641
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,2,1,power_law_1.2,0.6679449462890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,2,1,power_law_1.01,2.240037078857422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,2,1,power_law_1.01,7.263109741210937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,2,1,power_law_1.2,0.6948735809326172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,2,1,power_law_1.2,0.7264717102050782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,2,1,power_law_1.2,0.7531890869140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,2,1,power_law_1.2,0.7788409423828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,2,1,power_law_1.01,2.9175454711914064
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,2,1,power_law_1.2,0.8355289459228515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,2,1,power_law_1.2,1.1992633819580079
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,2,1,power_law_1.2,1.2329894256591796
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,2,1,power_law_1.2,1.2697676849365234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,2,1,power_law_1.2,1.3681465148925782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,2,1,power_law_1.2,1.4303385925292968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,2,1,power_law_1.2,1.5828282165527345
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,2,1,power_law_1.2,1.7120704650878906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,2,1,power_law_1.2,0.045368318557739255
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,2,1,power_law_1.2,2.063365173339844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,2,1,power_law_1.2,0.05962175846099853
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,2,1,power_law_1.2,0.09119872093200684
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,2,1,power_law_1.2,0.13662848472595215
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,2,1,power_law_1.2,0.18771392822265626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,2,1,power_law_1.2,0.23953279495239257
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,2,1,power_law_1.2,2.4537049865722658
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,2,1,power_law_1.2,0.34124992370605467
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,2,1,power_law_1.2,0.3590118408203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,2,1,power_law_1.2,0.3672902297973633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,2,1,power_law_1.2,0.38379070281982425
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,2,1,power_law_1.2,0.3935481643676758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,2,1,power_law_1.2,3.2152203369140624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,2,1,power_law_1.2,0.4058489608764648
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,2,1,power_law_1.2,0.4312998580932617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,2,1,power_law_1.2,0.4951436614990234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,2,1,power_law_1.2,0.5050892639160156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,2,1,power_law_1.2,0.5157478332519532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,2,1,power_law_1.2,4.117916870117187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,2,1,power_law_1.2,0.41599296569824223
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,2,1,power_law_1.2,0.5589363098144531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,2,1,power_law_1.2,0.5872812652587891
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,2,1,power_law_1.2,0.6691731262207031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,2,1,power_law_1.2,0.744645767211914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,2,1,power_law_1.2,0.8858732604980469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,2,1,power_law_1.2,1.059567337036133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,2,1,power_law_1.2,5.755987548828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,1,2,balanced,0.05005055904388428
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,1,2,balanced,0.06394239902496338
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,1,2,balanced,0.10820735931396483
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,2,1,power_law_1.2,1.3134867858886718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,1,2,balanced,0.35625343322753905
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,2,1,power_law_1.2,1.6646028137207032
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,1,2,balanced,0.49135486602783207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,1,2,balanced,0.48407806396484376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,1,2,balanced,0.4790643310546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,1,2,balanced,0.7113241577148438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,1,2,balanced,0.7122662353515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,1,2,balanced,0.7215923309326172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,1,2,balanced,0.7252031707763672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,2,1,power_law_1.2,2.4352633666992185
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,1,2,balanced,0.7410111999511718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,1,2,balanced,0.7536000061035156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,2,1,power_law_1.2,7.61083740234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,1,2,balanced,0.761181411743164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,1,2,balanced,0.7805452728271485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,1,2,balanced,0.8047577667236329
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,1,2,balanced,0.8446080017089844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,1,2,balanced,0.8736589050292969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,2,1,power_law_1.2,2.9150854492187497
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,1,2,balanced,0.9817292785644531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,1,2,balanced,1.0487667083740235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,1,2,balanced,0.21006975173950196
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,1,2,power_law_1.01,0.06283391952514648
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,1,2,balanced,1.4552230834960938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,1,2,power_law_1.01,0.0937497615814209
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,1,2,power_law_1.01,0.14713600158691406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,1,2,power_law_1.01,0.20552703857421878
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,1,2,power_law_1.01,0.2610380744934082
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,1,2,balanced,1.6462950134277343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,1,2,power_law_1.01,0.3830003356933594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,1,2,power_law_1.01,0.621899528503418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,1,2,power_law_1.01,0.6028121566772461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,1,2,balanced,2.461887969970703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,1,2,balanced,3.0303271484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,1,2,power_law_1.01,0.7178457641601563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,1,2,balanced,4.42995849609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,1,2,power_law_1.01,0.7331788635253906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,1,2,power_law_1.01,0.39666305541992186
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,1,2,power_law_1.01,0.40432769775390626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,1,2,power_law_1.01,0.7445977783203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,1,2,power_law_1.01,0.7997721862792969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,1,2,balanced,6.01048828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,1,2,power_law_1.01,0.8444057464599609
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,1,2,power_law_1.01,1.0458662414550781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,1,2,power_law_1.01,0.6736396789550781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,1,2,power_law_1.01,0.6433203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,1,2,power_law_1.01,0.6830233764648438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,1,2,power_law_1.01,1.5665664672851562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,1,2,power_law_1.2,0.065098237991333
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,1,2,power_law_1.2,0.08494848251342774
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,1,2,power_law_1.2,0.15631744384765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,1,2,power_law_1.01,2.6051480102539064
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,1,2,power_law_1.2,0.23527807235717774
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,1,2,power_law_1.2,0.3729087829589844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,1,2,power_law_1.01,0.9040332794189453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,1,2,power_law_1.01,3.532672119140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,1,2,power_law_1.2,0.40846977233886717
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,1,2,power_law_1.01,1.1709516906738282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,1,2,power_law_1.2,0.41520126342773434
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,1,2,power_law_1.2,0.5972620773315429
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,1,2,power_law_1.2,0.6398950576782226
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,1,2,power_law_1.2,0.667430419921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,1,2,power_law_1.2,0.684716796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,1,2,power_law_1.01,4.930549621582031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,1,2,power_law_1.01,1.8854988098144532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,1,2,power_law_1.2,0.7114854431152343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,1,2,power_law_1.2,0.7217971038818359
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,1,2,power_law_1.2,0.7471321868896484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,1,2,power_law_1.2,0.7405094146728516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,1,2,power_law_1.2,0.19503744125366212
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,1,2,power_law_1.2,0.7813785552978516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,1,2,power_law_1.2,0.846709747314453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,1,2,power_law_1.2,0.9058700561523437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,1,2,power_law_1.01,6.357947998046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,1,2,power_law_1.2,1.0766413116455078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,1,2,power_law_1.2,1.241338882446289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,1,2,power_law_1.2,1.6209869384765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,1,2,power_law_1.2,1.9368794250488282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,1,2,power_law_1.2,2.73707275390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,2,1,balanced,0.11478783607482909
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,1,2,power_law_1.2,3.7384179687500003
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,2,1,balanced,0.19996543884277346
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,2,1,balanced,0.06795904159545899
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,2,1,balanced,0.48503295898437504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,2,1,balanced,0.4924140930175781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,2,1,balanced,0.49359294891357425
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,2,1,balanced,0.36797374725341797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,2,1,balanced,0.5002540969848633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,2,1,balanced,0.5043987274169922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,2,1,balanced,0.49680831909179685
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,2,1,balanced,0.7253260803222656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,2,1,balanced,0.7363897705078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,2,1,balanced,0.7477516937255859
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,1,2,power_law_1.2,4.790868530273437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,2,1,balanced,0.7531603240966798
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,2,1,balanced,0.05031807899475098
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,2,1,balanced,0.7721440124511718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,2,1,balanced,0.7900077056884766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,2,1,balanced,0.8320480346679687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,2,1,balanced,0.8607315063476563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,1,2,power_law_1.2,6.805141601562499
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,2,1,balanced,0.9764006042480469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,2,1,balanced,1.0326988983154297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,2,1,power_law_1.01,0.05091392040252686
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,2,1,power_law_1.01,0.06958271980285644
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,2,1,power_law_1.01,0.11443967819213867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,2,1,balanced,1.3976237487792968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,2,1,balanced,1.5765592956542969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,2,1,power_law_1.01,0.4038547134399414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,2,1,balanced,2.350745544433594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,2,1,power_law_1.01,0.4174630355834961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,2,1,power_law_1.01,0.1878803253173828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,2,1,power_law_1.01,0.2542361640930176
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,2,1,power_law_1.01,0.43877502441406246
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,2,1,balanced,2.893436279296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,2,1,power_law_1.01,0.37736255645751954
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,2,1,power_law_1.01,0.44812671661376957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,2,1,power_law_1.01,0.6676729583740235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,2,1,power_law_1.01,0.6929612731933593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,2,1,power_law_1.01,0.722613754272461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,2,1,power_law_1.01,0.7305023956298828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,2,1,power_law_1.01,0.7355001831054688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,2,1,balanced,4.480862731933594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,2,1,power_law_1.01,0.7618323516845703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,2,1,power_law_1.01,0.8046137237548828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,2,1,power_law_1.01,0.45959617614746096
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,2,1,power_law_1.01,0.8538425445556641
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,2,1,power_law_1.01,0.9476441955566406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,2,1,power_law_1.01,1.0760665893554688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,2,1,balanced,5.463156127929688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,2,1,power_law_1.2,0.06943103790283203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,2,1,power_law_1.01,1.3678303527832032
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,2,1,power_law_1.2,0.11385472297668457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,2,1,power_law_1.01,1.6823648071289063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,2,1,power_law_1.2,0.37732223510742186
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,2,1,power_law_1.01,2.3550956726074217
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,2,1,power_law_1.2,0.40209087371826174
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,2,1,power_law_1.2,0.4283123016357422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,2,1,power_law_1.01,2.9480767822265626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,2,1,power_law_1.2,0.18252288818359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,2,1,power_law_1.2,0.44202495574951167
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,2,1,power_law_1.2,0.22561344146728515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,2,1,power_law_1.2,0.4549216079711914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,2,1,power_law_1.2,0.46819904327392575
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,2,1,power_law_1.2,0.6981452941894531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,2,1,power_law_1.2,0.7144294738769531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,2,1,power_law_1.01,4.243609008789063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,2,1,power_law_1.2,0.7291551971435547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,2,1,power_law_1.2,0.7380345916748047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,2,1,power_law_1.2,0.7549183654785157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,2,1,power_law_1.2,0.7770771026611328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,2,1,power_law_1.2,0.8163571166992188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,2,1,power_law_1.2,0.8523436737060546
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,2,1,power_law_1.2,0.9700198364257812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,2,1,power_law_1.2,1.1079334259033202
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,2,1,power_law_1.01,5.5365100097656255
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,2,1,power_law_1.2,1.40413818359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,1,2,balanced,0.39161087036132813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,2,1,power_law_1.2,1.7143801879882812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,2,1,power_law_1.2,0.05074431896209717
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,1,2,balanced,0.7256511688232422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,2,1,power_law_1.2,2.3422650146484374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,1,2,balanced,0.23198463439941408
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,1,2,balanced,2.6707647705078124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,2,1,power_law_1.2,2.980240783691406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,1,2,balanced,3.8816958618164064
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,1,2,balanced,1.4134080505371094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,1,2,balanced,3.9196493530273435
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,1,2,balanced,3.941529541015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,1,2,balanced,3.953893127441406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,2,1,power_law_1.2,4.2767813110351565
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,1,2,balanced,3.9587417602539063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,1,2,balanced,4.151553344726563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,1,2,balanced,4.056553039550781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,1,2,balanced,4.690037841796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,1,2,balanced,4.155584106445312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,2,1,power_law_1.2,5.587689819335937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,1,2,balanced,6.424422607421874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,1,2,balanced,6.615096435546874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,1,2,balanced,6.770863647460938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,1,2,balanced,7.222616577148438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,1,2,balanced,7.530122070312499
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,1,2,balanced,8.128734741210938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,1,2,balanced,0.12725119590759276
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,1,2,balanced,8.61984130859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,1,2,balanced,0.41724414825439454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,1,2,balanced,0.7672115325927734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,1,2,balanced,12.727596435546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,1,2,balanced,1.3506509399414062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,1,2,balanced,0.22878847122192383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,1,2,balanced,1.9611085510253907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,1,2,balanced,14.958148193359374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,1,2,balanced,1.9676914978027344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,1,2,balanced,2.0245350646972655
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,1,2,balanced,2.0057740783691407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,1,2,balanced,2.0317837524414064
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,1,2,balanced,2.0611724853515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,1,2,balanced,2.073322296142578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,1,2,balanced,2.1177357482910155
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,1,2,balanced,2.252711639404297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,1,2,balanced,23.965925292968752
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,1,2,balanced,2.3240870666503906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,1,2,balanced,1.97984130859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,1,2,balanced,2.3606643676757812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,1,2,balanced,2.536316223144531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,1,2,balanced,2.585118713378906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,1,2,balanced,2.764380187988281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,1,2,balanced,2.91973388671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,1,2,balanced,29.719724121093748
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,1,2,balanced,4.889666442871094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,1,2,power_law_1.01,0.38371585845947265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,1,2,balanced,5.130537109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,1,2,power_law_1.01,0.7006950378417969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,1,2,power_law_1.01,0.9447116851806641
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,1,2,power_law_1.01,1.5226573181152343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,1,2,balanced,7.40084228515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,1,2,balanced,7.90326416015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,1,2,power_law_1.01,1.9701222229003907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,1,2,power_law_1.01,3.04347900390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,1,2,power_law_1.01,3.1606719970703123
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,1,2,power_law_1.01,3.3613836669921873
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,1,2,power_law_1.01,3.3857281494140623
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,1,2,balanced,12.386466064453126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,1,2,power_law_1.01,3.481538696289063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,1,2,power_law_1.01,3.7763674926757815
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,1,2,power_law_1.01,3.94073974609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,1,2,power_law_1.01,4.547608337402344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,1,2,balanced,14.958930664062501
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,1,2,power_law_1.01,4.602900390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,1,2,balanced,44.696279296875005
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,1,2,power_law_1.01,6.052017822265625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,1,2,power_law_1.01,6.3196826171875005
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,1,2,power_law_1.01,6.529587402343751
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,1,2,power_law_1.01,7.08208740234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,1,2,power_law_1.01,7.668950805664062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,1,2,power_law_1.01,8.5181005859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,1,2,power_law_1.01,9.851383056640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,1,2,balanced,57.8027587890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,1,2,power_law_1.01,12.3799951171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,1,2,power_law_1.01,0.40178176879882815
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,1,2,power_law_1.01,0.5275980758666992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,1,2,power_law_1.01,0.7979853057861328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,1,2,power_law_1.01,0.9907826995849609
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,1,2,power_law_1.01,15.284317626953126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,1,2,power_law_1.01,1.529696044921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,1,2,power_law_1.01,0.2193484878540039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,1,2,power_law_1.01,1.585365753173828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,1,2,power_law_1.01,1.6774668884277344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,1,2,power_law_1.01,1.698357696533203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,1,2,power_law_1.01,1.738147888183594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,1,2,power_law_1.01,1.7961088562011718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,1,2,power_law_1.01,1.8192665100097656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,1,2,power_law_1.01,1.8608460998535157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,1,2,power_law_1.01,1.9040806579589844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,1,2,power_law_1.01,2.0665484619140626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,1,2,power_law_1.01,20.959650878906253
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,1,2,power_law_1.01,2.2072998046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,1,2,power_law_1.01,2.1958399963378907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,1,2,power_law_1.01,2.4021311950683595
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,1,2,power_law_1.01,2.637539978027344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,1,2,power_law_1.01,3.1333441162109374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,1,2,power_law_1.01,3.5310324096679686
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,1,2,power_law_1.01,4.5445184326171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,1,2,power_law_1.01,5.4424023437499995
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,1,2,power_law_1.01,27.18284912109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,1,2,power_law_1.2,0.38803199768066404
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,1,2,power_law_1.2,0.6089011383056641
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,1,2,power_law_1.2,0.9252620697021484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,1,2,power_law_1.01,7.256503295898438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,1,2,power_law_1.2,1.387298583984375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,1,2,power_law_1.2,1.8503948974609377
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,1,2,power_law_1.01,8.972316284179687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,1,2,power_law_1.2,3.0962316894531248
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,1,2,power_law_1.2,3.2629144287109377
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,1,2,power_law_1.2,3.3110552978515626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,1,2,power_law_1.2,3.4340045166015627
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,1,2,power_law_1.01,12.715447998046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,1,2,power_law_1.2,3.6145138549804683
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,1,2,power_law_1.2,3.8878195190429685
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,1,2,power_law_1.2,4.106117248535156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,1,2,power_law_1.01,39.987399902343746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,1,2,power_law_1.2,4.235990905761719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,1,2,power_law_1.01,16.524814453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,1,2,power_law_1.2,5.042703247070312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,1,2,power_law_1.2,6.274232177734374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,1,2,power_law_1.2,6.344031982421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,1,2,power_law_1.2,6.488961181640624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,1,2,power_law_1.2,7.3068695068359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,1,2,power_law_1.2,7.902137451171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,1,2,power_law_1.2,9.053604125976562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,1,2,power_law_1.01,52.167241210937505
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,1,2,power_law_1.2,10.359537353515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,1,2,power_law_1.2,0.22017663955688477
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,1,2,power_law_1.2,0.35550975799560547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,1,2,power_law_1.2,0.5394188690185546
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,1,2,power_law_1.2,0.744505615234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,1,2,power_law_1.2,0.9308313751220704
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,1,2,power_law_1.2,12.966301269531249
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,1,2,power_law_1.2,1.5517567443847657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,1,2,power_law_1.2,1.58136962890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,1,2,power_law_1.2,1.6683148193359376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,1,2,power_law_1.2,1.7053312683105468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,1,2,power_law_1.2,15.758574218749999
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,1,2,power_law_1.2,1.7374246215820315
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,1,2,power_law_1.2,1.7986534118652344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,1,2,power_law_1.2,1.8501210021972656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,1,2,power_law_1.2,1.9097984313964844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,1,2,power_law_1.2,2.014960632324219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,1,2,power_law_1.2,2.142142791748047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,1,2,power_law_1.2,2.2295884704589843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,1,2,power_law_1.2,2.245739593505859
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,1,2,power_law_1.2,2.544451904296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,1,2,power_law_1.2,2.703111572265625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,1,2,power_law_1.2,22.52838134765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,1,2,power_law_1.2,3.166069641113281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,1,2,power_law_1.2,3.633395080566406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,1,2,power_law_1.2,4.6794482421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,1,2,power_law_1.2,5.689666748046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,2,1,balanced,0.26344192504882813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,2,1,balanced,0.3971923065185547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,1,2,power_law_1.2,29.95401123046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,1,2,power_law_1.2,7.369896850585937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,2,1,balanced,0.7195270538330079
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,2,1,balanced,1.3615603637695313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,1,2,power_law_1.2,9.303271484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,2,1,balanced,3.9104467773437497
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,2,1,balanced,3.9529003906249995
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,2,1,balanced,2.654479370117188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,1,2,power_law_1.2,13.094316406249998
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,2,1,balanced,3.978035888671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,2,1,balanced,4.290006408691406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,1,2,power_law_1.2,41.4725634765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,1,2,power_law_1.2,16.73582763671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,2,1,balanced,4.044858093261719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,2,1,balanced,4.027830505371094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,2,1,balanced,4.077090454101563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,2,1,balanced,4.16065673828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,2,1,balanced,4.151627502441406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,2,1,balanced,4.2057363891601565
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,2,1,balanced,5.41103759765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,2,1,balanced,6.5187451171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,2,1,balanced,7.098833618164062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,2,1,balanced,7.420456542968751
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,2,1,balanced,8.220149536132812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,2,1,balanced,8.793941650390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,1,2,power_law_1.2,55.227690429687506
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,2,1,balanced,0.14538240432739258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,2,1,balanced,0.24418367385864256
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,2,1,balanced,12.143814697265626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,2,1,balanced,0.4276403045654297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,2,1,balanced,13.616903076171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,2,1,balanced,0.7388678741455078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,2,1,balanced,1.3627731323242187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,2,1,balanced,2.014072265625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,2,1,balanced,2.0149920654296873
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,2,1,balanced,2.070112609863281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,2,1,balanced,2.029605712890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,2,1,balanced,2.0502726745605466
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,2,1,balanced,2.087671051025391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,2,1,balanced,2.1119961547851562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,2,1,balanced,20.93787841796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,2,1,balanced,2.124680938720703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,2,1,balanced,2.1484249877929686
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,2,1,balanced,2.164853057861328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,2,1,balanced,2.3063897705078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,2,1,balanced,2.2207513427734376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,2,1,balanced,27.503889160156252
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,2,1,balanced,2.6296792602539063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,2,1,balanced,2.700703430175781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,2,1,balanced,2.8933306884765626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,2,1,balanced,3.0876498413085938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,2,1,balanced,4.802272033691406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,2,1,balanced,7.998259887695312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,2,1,balanced,5.244407958984375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,2,1,balanced,8.573598022460938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,2,1,power_law_1.01,0.7216172790527343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,2,1,power_law_1.01,0.2562892723083496
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,2,1,power_law_1.01,0.39386688232421874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,2,1,power_law_1.01,1.7605183410644532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,2,1,balanced,41.4950634765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,2,1,power_law_1.01,2.8801977539062498
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,2,1,power_law_1.01,1.2365087890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,2,1,balanced,13.109222412109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,2,1,power_law_1.01,3.0957620239257815
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,2,1,power_law_1.01,3.255585327148437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,2,1,power_law_1.01,3.3573709106445313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,2,1,power_law_1.01,3.401436157226563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,2,1,power_law_1.01,3.551705627441406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,2,1,balanced,16.215323486328124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,2,1,power_law_1.01,3.9387603759765626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,2,1,power_law_1.01,4.275564880371094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,2,1,power_law_1.01,4.618880615234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,2,1,power_law_1.01,4.9664282226562495
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,2,1,power_law_1.01,5.29228515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,2,1,power_law_1.01,6.775009765625001
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,2,1,power_law_1.01,6.834696655273437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,2,1,balanced,53.3945458984375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,2,1,power_law_1.01,7.310511474609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,2,1,power_law_1.01,8.147702026367188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,2,1,power_law_1.01,9.246407470703124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,2,1,power_law_1.01,0.14150464057922363
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,2,1,power_law_1.01,0.4271379089355468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,2,1,power_law_1.01,11.716314697265625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,2,1,power_law_1.01,0.24097728729248047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,2,1,power_law_1.01,14.423961181640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,2,1,power_law_1.01,0.6747833251953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,2,1,power_law_1.01,1.5954138183593751
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,2,1,power_law_1.01,1.4878854370117187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,2,1,power_law_1.01,0.9189778900146484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,2,1,power_law_1.01,1.7342182922363283
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,2,1,power_law_1.01,1.7578515625000002
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,2,1,power_law_1.01,1.676629180908203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,2,1,power_law_1.01,1.80350341796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,2,1,power_law_1.01,19.57639404296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,2,1,power_law_1.01,1.8631584167480468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,2,1,power_law_1.01,1.8848985290527345
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,2,1,power_law_1.01,1.9641920471191405
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,2,1,power_law_1.01,1.9787059020996094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,2,1,power_law_1.01,2.055043182373047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,2,1,power_law_1.01,2.1732524108886717
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,2,1,power_law_1.01,2.7938201904296873
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,2,1,power_law_1.01,3.179959716796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,2,1,power_law_1.01,25.05609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,2,1,power_law_1.01,3.542155456542969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,2,1,power_law_1.01,4.620814208984375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,2,1,power_law_1.01,2.580979919433594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,2,1,power_law_1.2,0.26199615478515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,2,1,power_law_1.2,0.3919558334350586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,2,1,power_law_1.01,5.499594116210938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,2,1,power_law_1.2,0.7208134460449218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,2,1,power_law_1.2,1.1400000000000001
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,2,1,power_law_1.01,7.125546264648437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,2,1,power_law_1.2,1.5565055847167968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,2,1,power_law_1.01,9.281724243164062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,2,1,power_law_1.2,2.839796447753906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,2,1,power_law_1.2,3.0860491943359376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,2,1,power_law_1.01,35.53381591796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,2,1,power_law_1.2,3.196759033203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,2,1,power_law_1.01,12.75228271484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,2,1,power_law_1.2,3.356486511230469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,2,1,power_law_1.2,3.4912646484375003
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,2,1,power_law_1.2,3.6379461669921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,2,1,power_law_1.2,4.034657897949218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,2,1,power_law_1.2,4.415863647460937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,2,1,power_law_1.2,4.507096252441406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,2,1,power_law_1.01,16.401614990234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,2,1,power_law_1.2,5.525780029296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,2,1,power_law_1.2,5.7249517822265625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,2,1,power_law_1.2,6.436782226562499
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,2,1,power_law_1.01,46.4829150390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,2,1,power_law_1.2,7.097090454101563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,2,1,power_law_1.2,7.4286303710937505
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,2,1,power_law_1.2,8.397072753906249
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,2,1,power_law_1.2,9.617304077148438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,2,1,power_law_1.2,0.24097215652465823
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,2,1,power_law_1.2,0.4281689453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,2,1,power_law_1.2,0.141843843460083
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,2,1,power_law_1.2,12.209140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,2,1,power_law_1.2,0.6265689468383788
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,2,1,power_law_1.2,0.8315500640869141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,2,1,power_law_1.2,1.4698713684082032
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,2,1,power_law_1.2,1.6032333374023438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,2,1,power_law_1.2,1.6575296020507815
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,2,1,power_law_1.2,14.70135498046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,2,1,power_law_1.2,1.7344595336914062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,2,1,power_law_1.2,1.7853868103027346
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,2,1,power_law_1.2,1.8372166442871094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,2,1,power_law_1.2,1.8956640624999999
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,2,1,power_law_1.2,1.9421197509765626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,2,1,power_law_1.2,2.0402195739746096
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,2,1,power_law_1.2,2.079621124267578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,2,1,power_law_1.2,2.1446099853515626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,2,1,power_law_1.2,2.2836831665039066
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,2,1,power_law_1.2,19.760538330078127
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,2,1,power_law_1.2,2.7198214721679688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,2,1,power_law_1.2,2.8851885986328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,2,1,power_law_1.2,3.3467098999023435
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,2,1,power_law_1.2,3.858948974609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,2,1,power_law_1.2,4.541761779785157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,2,1,power_law_1.2,5.66095947265625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,1,4,balanced,0.20655744552612304
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,2,1,power_law_1.2,25.31077880859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,1,4,balanced,0.1831065559387207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,1,4,balanced,0.3774950408935547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,2,1,power_law_1.2,7.244401245117187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,1,4,balanced,0.41830783843994135
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,1,4,balanced,0.38092670440673826
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,1,4,balanced,0.29992576599121096
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,1,4,balanced,0.382977294921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,1,4,balanced,0.38208129882812497
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,1,4,balanced,0.3887526321411133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,1,4,balanced,0.38974590301513673
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,1,4,balanced,0.3980147171020508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,1,4,balanced,0.4083212661743164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,1,4,balanced,0.4196646499633789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,1,4,balanced,0.4306393432617187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,1,4,balanced,0.46032768249511713
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,1,4,balanced,0.46951038360595704
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,1,4,balanced,0.4935129547119141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,1,4,balanced,0.6580236816406251
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,1,4,balanced,0.7226201629638671
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,1,4,balanced,1.1384217834472656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,2,1,power_law_1.2,9.540023803710938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,1,4,balanced,1.4218751525878905
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,1,4,balanced,2.141558990478516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,1,4,balanced,2.7968896484375003
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,1,4,balanced,4.155691528320313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,1,4,balanced,0.0956339168548584
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,1,4,balanced,0.09320704460144043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,1,4,balanced,0.13333888053894044
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,1,4,balanced,0.161474552154541
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,1,4,balanced,0.16235008239746093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,1,4,balanced,5.574492187500001
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,1,4,balanced,0.17244672775268555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,1,4,balanced,0.1730931282043457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,1,4,balanced,0.1696895980834961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,1,4,balanced,0.1672755241394043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,1,4,balanced,0.16294784545898439
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,1,4,balanced,0.16855552673339844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,1,4,balanced,0.1702822494506836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,1,4,balanced,0.17217664718627929
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,1,4,balanced,0.175097599029541
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,1,4,balanced,0.1797222328186035
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,1,4,balanced,0.182871036529541
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,1,4,balanced,0.19088895797729494
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,1,4,balanced,0.23519359588623048
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,1,4,balanced,0.2362611198425293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,1,4,balanced,0.30083711624145504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,1,4,balanced,0.36689151763916017
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,1,4,balanced,0.5565299224853516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,1,4,balanced,0.6997427368164063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,1,4,balanced,9.705307006835938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,1,4,balanced,1.1023193359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,1,4,balanced,1.49119873046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,1,4,power_law_1.01,0.18764799118041992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,1,4,power_law_1.01,0.2840396881103516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,1,4,balanced,2.292376251220703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,1,4,power_law_1.01,0.3390707015991211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,2,1,power_law_1.2,36.5200634765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,2,1,power_law_1.2,13.209239501953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,1,4,power_law_1.01,0.3854886245727539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,1,4,power_law_1.01,0.37604095458984377
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,1,4,power_law_1.01,0.44531200408935545
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,1,4,balanced,2.936408386230469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,1,4,power_law_1.01,0.42955646514892576
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,1,4,power_law_1.01,0.4088729476928711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,1,4,power_law_1.01,0.4129011154174805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,1,4,power_law_1.01,0.42582145690917966
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,1,4,power_law_1.01,0.4506291198730469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,1,4,power_law_1.01,0.4574067306518555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,1,4,power_law_1.01,0.4717030334472656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,1,4,power_law_1.01,0.5344614410400391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,1,4,power_law_1.01,0.6265241622924804
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,1,4,balanced,12.709483642578125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,1,4,power_law_1.01,0.7185906982421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,1,4,power_law_1.01,0.8424332427978516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,1,4,power_law_1.01,1.28384765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,1,4,power_law_1.01,1.7038003540039064
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,1,4,power_law_1.01,2.4770367431640627
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,1,4,power_law_1.01,2.681299133300781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,1,4,power_law_1.01,5.201860961914063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,1,4,power_law_1.01,0.0956287956237793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,1,4,power_law_1.01,0.12610048294067383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,1,4,power_law_1.01,0.16275583267211916
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,1,4,power_law_1.01,0.16167167663574217
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,1,4,power_law_1.01,0.16260351181030272
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,1,4,power_law_1.01,0.16423423767089845
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,1,4,power_law_1.01,0.16604032516479492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,1,4,power_law_1.01,0.16742143630981446
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,1,4,power_law_1.01,0.16892799377441406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,1,4,power_law_1.01,0.17101696014404297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,1,4,power_law_1.01,0.17387519836425783
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,1,4,power_law_1.01,0.18265087127685548
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,1,4,power_law_1.01,0.18789119720458985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,1,4,power_law_1.01,0.19655040740966795
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,1,4,power_law_1.01,0.21689599990844727
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,1,4,power_law_1.01,0.23147903442382814
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,1,4,power_law_1.01,0.26340095520019535
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,2,1,power_law_1.2,16.280147705078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,1,4,power_law_1.01,0.3544947052001953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,1,4,power_law_1.01,5.877640991210937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,1,4,power_law_1.01,0.43819519042968746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,1,4,power_law_1.01,0.5626163101196289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,1,4,power_law_1.01,0.7020877075195313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,1,4,power_law_1.01,8.869288940429687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,1,4,power_law_1.01,1.589580841064453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,1,4,power_law_1.01,1.1781887817382812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,2,1,power_law_1.2,47.264003906250004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,1,4,power_law_1.2,0.1917862319946289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,1,4,power_law_1.01,14.356361083984376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,1,4,power_law_1.2,0.36573310852050783
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,1,4,power_law_1.2,0.3752614212036133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,1,4,power_law_1.2,0.37631744384765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,1,4,power_law_1.2,0.23654527664184571
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,1,4,power_law_1.2,0.38186752319335937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,1,4,power_law_1.2,0.3890854263305664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,1,4,power_law_1.2,0.3989734268188477
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,1,4,power_law_1.2,0.41248001098632814
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,1,4,power_law_1.2,0.4282675170898438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,1,4,power_law_1.2,0.45110145568847654
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,1,4,power_law_1.2,0.4556595230102539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,1,4,power_law_1.01,2.1817280578613283
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,1,4,power_law_1.2,0.46953601837158204
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,1,4,power_law_1.2,0.540079345703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,1,4,power_law_1.2,0.6347366333007812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,1,4,power_law_1.2,0.7034559631347657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,1,4,power_law_1.01,3.179864196777344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,1,4,power_law_1.2,0.841427230834961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,1,4,power_law_1.2,1.3038873291015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,1,4,power_law_1.2,1.6824972534179685
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,1,4,power_law_1.2,2.2452838134765623
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,1,4,power_law_1.01,5.229004516601562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,1,4,power_law_1.2,2.702893981933594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,1,4,power_law_1.2,4.635166625976562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,1,4,power_law_1.01,21.412663574218747
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,1,4,power_law_1.2,0.09217920303344726
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,1,4,power_law_1.2,0.1325235176086426
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,1,4,power_law_1.2,0.1562662410736084
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,1,4,power_law_1.2,0.1621887969970703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,1,4,power_law_1.2,0.1630963134765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,1,4,power_law_1.2,0.16503168106079102
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,1,4,power_law_1.2,0.16663551330566406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,1,4,power_law_1.2,0.16788095474243164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,1,4,power_law_1.2,0.1700351905822754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,1,4,power_law_1.2,0.1715635108947754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,1,4,power_law_1.2,0.17477760314941407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,1,4,power_law_1.2,0.1866201591491699
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,1,4,power_law_1.2,0.188986873626709
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,1,4,power_law_1.2,0.19738239288330078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,1,4,power_law_1.2,0.21979135513305664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,1,4,power_law_1.2,0.231276798248291
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,1,4,power_law_1.2,0.2635686492919922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,1,4,power_law_1.01,6.276906127929688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,1,4,power_law_1.2,5.7774157714843755
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,1,4,power_law_1.2,0.3780979156494141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,1,4,power_law_1.2,0.4653299331665039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,1,4,power_law_1.2,0.6294988632202149
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,1,4,power_law_1.2,0.8864908599853516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,1,4,power_law_1.2,8.981354370117188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,1,4,power_law_1.2,1.166792984008789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,1,4,power_law_1.2,13.74237548828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,1,4,power_law_1.2,2.584927978515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,2,2,balanced,0.09766400337219239
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,1,4,power_law_1.01,39.72623291015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,2,2,balanced,0.12325887680053711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,2,2,balanced,0.29517440795898436
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,2,2,balanced,0.29639936447143556
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,2,2,balanced,0.2977420806884766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,2,2,balanced,0.30102655410766604
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,2,2,balanced,0.3033216094970703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,2,2,balanced,0.30715776443481446
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,2,2,balanced,0.3119590377807617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,2,2,balanced,0.31463808059692383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,2,2,balanced,0.3288051223754883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,2,2,balanced,0.23741184234619142
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,2,2,balanced,0.3349964904785156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,2,2,balanced,0.3546316909790039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,2,2,balanced,0.4040192031860352
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,2,2,balanced,0.4018880081176758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,2,2,balanced,0.42072704315185544
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,2,2,balanced,0.6130099105834961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,2,2,balanced,0.7728038024902344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,1,4,power_law_1.2,1.5632205200195313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,2,2,balanced,1.1292646026611328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,2,2,balanced,1.4032025146484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,2,2,balanced,2.1080934143066408
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,2,2,balanced,2.713182678222656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,2,2,balanced,4.054401245117187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,2,2,balanced,5.427492065429687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,2,2,balanced,0.05815807819366455
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,2,2,balanced,0.07515007972717284
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,2,2,balanced,9.280855102539062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,2,2,balanced,0.12914175987243653
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,2,2,balanced,0.12835968017578125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,2,2,balanced,0.12897279739379883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,2,2,balanced,0.12894975662231445
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,1,4,power_law_1.2,3.704266357421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,2,2,balanced,0.13045120239257812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,2,2,balanced,0.1377945613861084
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,2,2,balanced,0.13241727828979494
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,2,2,balanced,0.13760128021240234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,2,2,balanced,0.13799167633056642
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,2,2,balanced,0.14227328300476075
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,2,2,balanced,0.14721792221069335
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,2,2,balanced,0.1542527961730957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,2,2,balanced,0.1590707206726074
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,2,2,balanced,0.15767807960510255
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,2,2,balanced,0.115033597946167
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,2,2,balanced,0.20615167617797853
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,2,2,balanced,0.22420223236083983
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,2,2,balanced,0.3158092880249023
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,2,2,balanced,0.3801689529418945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,2,2,balanced,0.5910374450683593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,2,2,balanced,0.7383897399902344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,2,2,balanced,12.156094970703126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,2,2,balanced,1.1718450927734376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,2,2,power_law_1.01,0.11554176330566406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,2,2,power_law_1.01,0.16517887115478516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,2,2,balanced,1.5230592346191405
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,2,2,power_law_1.01,0.18801408767700195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,2,2,power_law_1.01,0.30183935165405273
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,2,2,power_law_1.01,0.29622272491455076
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,1,4,power_law_1.2,21.728581542968747
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,2,2,power_law_1.01,0.30161792755126954
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,2,2,power_law_1.01,0.3045465660095215
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,2,2,power_law_1.01,0.3156863975524902
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,2,2,balanced,2.422137603759766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,2,2,power_law_1.01,0.32504959106445314
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,2,2,power_law_1.01,0.33307647705078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,2,2,power_law_1.01,0.3799935913085938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,2,2,power_law_1.01,0.3931315231323242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,2,2,power_law_1.01,0.40112640380859377
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,2,2,balanced,3.0619161987304686
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,2,2,power_law_1.01,0.4154777526855469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,2,2,power_law_1.01,0.4504537582397461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,2,2,power_law_1.01,0.46381183624267575
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,2,2,power_law_1.01,0.5868851089477539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,2,2,power_law_1.01,0.966961898803711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,2,2,power_law_1.01,1.1451929473876954
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,2,2,power_law_1.01,1.4275494384765626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,2,2,power_law_1.01,1.641148223876953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,1,4,power_law_1.2,5.0025765991210935
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,2,2,power_law_1.01,3.2368283081054683
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,2,2,power_law_1.01,0.0686079978942871
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,2,2,power_law_1.01,0.07983232021331788
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,2,2,power_law_1.01,0.09699456214904785
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,2,2,power_law_1.01,0.1245081615447998
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,2,2,power_law_1.01,0.12519424438476562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,2,2,power_law_1.01,0.12676480293273926
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,2,2,power_law_1.01,0.128021764755249
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,2,2,power_law_1.01,0.12890751838684084
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,2,2,power_law_1.01,0.13048319816589354
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,2,2,power_law_1.01,0.1312332820892334
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,2,2,power_law_1.01,0.13430527687072752
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,2,2,power_law_1.01,0.1439743995666504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,2,2,power_law_1.01,0.1470400047302246
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,2,2,power_law_1.01,0.15861760139465333
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,2,2,power_law_1.01,0.1665433692932129
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,2,2,power_law_1.01,0.1712985610961914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,2,2,power_law_1.01,5.348016357421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,2,2,power_law_1.01,0.1856780815124512
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,2,2,power_law_1.01,0.28910848617553714
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,2,2,power_law_1.01,0.249171199798584
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,2,2,power_law_1.01,0.5540326309204102
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,2,2,power_law_1.01,0.4071257781982422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,2,2,power_law_1.01,4.114167175292969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,1,4,power_law_1.2,7.428601684570313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,2,2,power_law_1.01,1.1444671630859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,2,2,power_law_1.01,1.4069734191894532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,2,2,power_law_1.01,0.8300262451171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,2,2,power_law_1.2,0.11513343811035157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,1,4,power_law_1.2,34.570156250000004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,2,2,power_law_1.2,0.16575231552124023
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,2,2,power_law_1.2,0.2956454467773438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,2,2,power_law_1.2,0.2956492805480957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,2,2,power_law_1.2,0.29966207504272463
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,2,2,power_law_1.2,0.30564992904663085
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,2,2,power_law_1.2,0.3102387237548828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,2,2,power_law_1.2,0.3248345565795899
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,2,2,power_law_1.2,0.3423680114746094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,2,2,power_law_1.01,8.113466796874999
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,2,2,power_law_1.2,0.19356927871704102
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,2,2,power_law_1.2,0.37575038909912106
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,2,2,power_law_1.2,0.378284797668457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,2,2,power_law_1.2,0.40029182434082033
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,2,2,power_law_1.2,0.41765502929687504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,2,2,power_law_1.2,0.4517990493774414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,2,2,power_law_1.2,0.45419776916503907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,2,2,power_law_1.2,0.5946803283691406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,2,2,power_law_1.2,0.8569062042236328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,2,2,power_law_1.01,2.0910745239257813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,2,2,power_law_1.2,1.1144895935058594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,2,2,power_law_1.2,1.5586956787109376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,2,2,power_law_1.2,1.6212556457519534
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,2,2,power_law_1.01,12.574826660156251
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,2,2,power_law_1.2,3.4361038208007812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,2,2,power_law_1.01,3.012628479003906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,2,2,power_law_1.2,0.06894336223602296
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,2,2,power_law_1.2,5.6511383056640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,2,2,power_law_1.2,0.08076671600341798
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,2,2,power_law_1.2,0.10716544151306154
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,2,2,power_law_1.2,0.1263219165802002
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,2,2,power_law_1.2,0.12755071640014648
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,2,2,power_law_1.2,0.12911104202270507
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,2,2,power_law_1.2,0.1303014373779297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,2,2,power_law_1.2,0.13011839866638183
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,2,2,power_law_1.2,0.13049471855163575
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,2,2,power_law_1.2,0.1320524787902832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,2,2,power_law_1.01,19.252446289062497
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,2,2,power_law_1.2,0.13464320182800293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,2,2,power_law_1.2,0.14197759628295897
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,2,2,power_law_1.2,3.8979058837890626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,2,2,power_law_1.2,0.14753151893615724
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,2,2,power_law_1.2,0.16299520492553712
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,2,2,power_law_1.2,0.1722355270385742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,2,2,power_law_1.2,0.16771583557128905
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,2,2,power_law_1.2,0.19250816345214844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,2,2,power_law_1.2,0.2728934478759766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,2,2,power_law_1.2,0.299686393737793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,2,2,power_law_1.2,0.39729663848876956
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,2,2,power_law_1.01,4.685035400390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,2,2,power_law_1.2,0.5590464019775391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,2,2,power_law_1.2,1.1077606201171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,2,2,power_law_1.2,7.906358032226562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,2,2,power_law_1.2,0.8408678436279298
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,2,2,power_law_1.2,1.5666111755371095
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,4,1,balanced,0.0661836814880371
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,4,1,balanced,0.1934227180480957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,4,1,balanced,0.10758912086486816
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,2,2,power_law_1.2,2.1333351135253906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,4,1,balanced,0.1925779151916504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,4,1,balanced,0.20774335861206056
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,4,1,balanced,0.21042688369750975
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,2,2,power_law_1.2,3.1479922485351564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,4,1,balanced,0.19359167098999025
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,4,1,balanced,0.19846656799316406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,4,1,balanced,0.2039200019836426
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,4,1,balanced,0.20074432373046874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,4,1,balanced,0.2236185646057129
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,4,1,balanced,0.20040767669677734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,4,1,balanced,0.2317888069152832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,4,1,balanced,0.23114303588867186
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,4,1,balanced,0.334582405090332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,4,1,balanced,0.5025478363037109
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,4,1,balanced,0.6437229156494141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,4,1,balanced,0.20124544143676756
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,4,1,balanced,0.9815904235839843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,4,1,balanced,0.20373567581176757
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,4,1,balanced,1.2858265686035157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,2,2,power_law_1.2,13.48845458984375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,4,1,balanced,1.9225088500976562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,4,1,balanced,3.7775213623046873
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,4,1,balanced,0.22618175506591798
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,4,1,balanced,0.3467737579345703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,4,1,balanced,5.095542907714844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,4,1,balanced,0.04881472110748291
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,4,1,balanced,0.0709990406036377
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,4,1,balanced,0.12253952026367188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,4,1,balanced,2.5638943481445313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,4,1,balanced,0.11459391593933106
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,4,1,balanced,0.1181657600402832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,4,1,balanced,0.12050111770629883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,4,1,balanced,0.12153663635253906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,4,1,balanced,0.12337599754333498
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,4,1,balanced,0.12526911735534668
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,2,2,power_law_1.2,4.581661376953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,4,1,balanced,0.1685536003112793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,4,1,balanced,0.12362879753112792
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,4,1,balanced,0.14775424003601073
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,4,1,balanced,0.1802016067504883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,4,1,balanced,0.20094015121459963
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,4,1,balanced,0.2141017532348633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,4,1,balanced,0.358485107421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,4,1,balanced,0.5407731246948242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,4,1,balanced,0.7207148742675782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,4,1,balanced,0.11675135612487793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,4,1,balanced,1.1038246154785156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,2,2,power_law_1.2,17.835997314453124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,4,1,balanced,1.4400831604003907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,4,1,balanced,0.11913472175598146
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,4,1,balanced,2.1516607666015624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,4,1,balanced,0.11772352218627931
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,4,1,balanced,2.885285034179687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,4,1,balanced,0.13742976188659667
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,4,1,power_law_1.01,0.21494911193847654
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,4,1,power_law_1.01,0.1335251235961914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,4,1,power_law_1.01,0.19637184143066405
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,4,1,power_law_1.01,0.10383744239807129
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,4,1,power_law_1.01,0.06657792091369628
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,4,1,power_law_1.01,0.2296352005004883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,4,1,power_law_1.01,0.2059481620788574
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,4,1,power_law_1.01,0.25757312774658203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,4,1,power_law_1.01,0.27453376770019533
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,4,1,power_law_1.01,0.18897344589233397
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,4,1,power_law_1.01,0.2820364761352539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,4,1,power_law_1.01,0.27641599655151367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,4,1,power_law_1.01,0.27835199356079104
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,4,1,balanced,0.2955673599243164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,4,1,power_law_1.01,0.3739231872558594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,4,1,power_law_1.01,0.42389759063720706
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,4,1,power_law_1.01,0.19508031845092771
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,4,1,power_law_1.01,0.5777542495727539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,4,1,power_law_1.01,0.1934502410888672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,4,1,power_law_1.01,0.6942176055908204
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,4,1,power_law_1.01,0.24472383499145506
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,4,1,power_law_1.01,1.0730976104736327
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,4,1,power_law_1.01,0.20500864028930663
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,4,1,power_law_1.01,0.06862463951110839
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,4,1,power_law_1.01,0.11961791992187501
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,4,1,power_law_1.01,1.3066476440429686
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,4,1,power_law_1.01,0.1147430419921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,4,1,power_law_1.01,0.11740736007690429
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,4,1,power_law_1.01,0.11786687850952149
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,4,1,power_law_1.01,0.11765248298645019
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,4,1,power_law_1.01,0.11787199974060059
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,4,1,power_law_1.01,0.12326848030090332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,4,1,power_law_1.01,0.13527615547180177
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,4,1,power_law_1.01,0.14318528175354003
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,4,1,power_law_1.01,0.1522111988067627
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,4,1,power_law_1.01,0.14510144233703612
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,4,1,power_law_1.01,0.1624166488647461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,4,1,power_law_1.01,0.16945791244506836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,4,1,power_law_1.01,0.16748735427856445
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,4,1,power_law_1.01,0.21606143951416015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,4,1,power_law_1.01,0.24639423370361327
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,4,1,power_law_1.01,2.0058937072753906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,4,1,power_law_1.01,0.33023361206054686
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,4,1,power_law_1.01,0.04912511825561523
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,4,1,power_law_1.01,0.40861694335937504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,4,1,power_law_1.01,2.574132995605469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,4,1,power_law_1.01,0.0807852840423584
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,4,1,power_law_1.01,0.5842943954467773
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,4,1,power_law_1.2,0.06846720218658447
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,4,1,power_law_1.2,0.10038208007812499
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,4,1,power_law_1.01,0.7253990173339844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,4,1,power_law_1.2,0.11507391929626465
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,4,1,power_law_1.01,1.0949311828613282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,4,1,power_law_1.01,3.9073150634765623
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,4,1,power_law_1.2,0.18772031784057616
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,4,1,power_law_1.2,0.19492479324340822
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,4,1,power_law_1.01,1.4964031982421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,4,1,power_law_1.2,0.19158016204833986
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,4,1,power_law_1.2,0.21198911666870118
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,4,1,power_law_1.2,0.250696964263916
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,4,1,power_law_1.2,0.25690303802490233
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,4,1,power_law_1.2,0.2758489608764648
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,4,1,power_law_1.2,0.2918521690368652
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,4,1,power_law_1.2,0.28376447677612304
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,4,1,power_law_1.2,0.19550207138061523
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,4,1,power_law_1.2,0.19972864151000977
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,4,1,power_law_1.2,0.2791033554077148
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,4,1,power_law_1.2,0.23585344314575196
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,4,1,power_law_1.2,0.3627289581298828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,4,1,power_law_1.2,0.20756544113159178
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,4,1,power_law_1.01,2.199096984863281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,4,1,power_law_1.01,5.159894409179687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,4,1,power_law_1.2,0.43486785888671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,4,1,power_law_1.2,0.5663641738891602
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,4,1,power_law_1.2,0.732437744140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,4,1,power_law_1.2,1.0376338958740234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,4,1,power_law_1.2,0.068220157623291
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,4,1,power_law_1.01,2.9749368286132816
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,4,1,power_law_1.2,0.12350784301757814
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,4,1,power_law_1.2,0.11341055870056152
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,4,1,power_law_1.2,0.04939455986022949
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,4,1,power_law_1.2,0.07614272117614747
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,4,1,power_law_1.2,0.11721856117248534
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,4,1,power_law_1.2,0.11885503768920898
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,4,1,power_law_1.2,1.9939987182617187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,4,1,power_law_1.2,0.1238924789428711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,4,1,power_law_1.2,0.14517120361328123
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,4,1,power_law_1.2,0.11639616012573242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,4,1,power_law_1.2,0.15096768379211425
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,4,1,power_law_1.2,0.14352191925048827
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,4,1,power_law_1.2,0.1576416015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,4,1,power_law_1.2,0.16876480102539063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,4,1,power_law_1.2,0.16713024139404298
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,4,1,power_law_1.2,0.21149887084960936
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,4,1,power_law_1.2,0.11785920143127442
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,4,1,power_law_1.2,1.3866969299316405
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,4,1,power_law_1.2,0.2438355255126953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,4,1,power_law_1.2,0.1377286434173584
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,4,1,power_law_1.2,0.31914047241210936
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,4,1,power_law_1.2,2.6484103393554688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,4,1,power_law_1.2,0.4050438308715821
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,4,1,power_law_1.2,0.5716230392456054
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,1,4,balanced,0.2404275131225586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,1,4,balanced,0.23504640579223635
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,1,4,balanced,0.39403137207031247
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,1,4,balanced,0.5589836883544922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,1,4,balanced,0.5616320037841798
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,1,4,balanced,0.565797119140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,4,1,power_law_1.2,0.7596691131591797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,1,4,balanced,0.5742208099365234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,1,4,balanced,0.5847974395751953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,1,4,balanced,0.5872256088256835
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,1,4,balanced,0.5917657470703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,1,4,balanced,0.5991257476806641
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,1,4,balanced,0.624832000732422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,1,4,balanced,0.6301055908203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,1,4,balanced,0.6314201736450196
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,1,4,balanced,0.6852722930908203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,1,4,balanced,0.6982617950439453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,1,4,balanced,0.7408499145507813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,1,4,balanced,1.1286720275878905
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,4,1,power_law_1.2,1.111785583496094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,1,4,balanced,1.3186009216308594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,1,4,balanced,1.9386341857910157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,4,1,power_law_1.2,3.887911682128906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,1,4,balanced,2.4174310302734376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,1,4,balanced,3.63736572265625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,1,4,balanced,4.6594738769531245
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,4,1,power_law_1.2,1.486691131591797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,1,4,balanced,0.12126336097717286
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,1,4,balanced,0.11997695922851563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,1,4,balanced,7.457457885742187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,1,4,balanced,0.23826175689697265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,1,4,balanced,0.22288000106811526
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,1,4,balanced,0.21806720733642576
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,1,4,balanced,0.21791872024536132
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,1,4,balanced,0.21803775787353516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,1,4,balanced,0.21880191802978516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,1,4,balanced,0.21961599349975586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,1,4,balanced,0.22150527954101565
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,1,4,balanced,0.22445823669433596
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,1,4,balanced,9.657432250976562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,1,4,balanced,0.22663936614990235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,1,4,balanced,0.2485798454284668
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,1,4,balanced,0.23656448364257812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,1,4,balanced,0.24382207870483397
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,1,4,balanced,0.25268480300903323
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,1,4,balanced,0.31429248809814453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,1,4,balanced,0.33294208526611324
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,1,4,balanced,0.5021900939941406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,1,4,balanced,0.6203545761108399
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,1,4,balanced,0.8816230773925782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,1,4,balanced,1.1977049255371095
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,1,4,balanced,0.18746623992919922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,1,4,balanced,1.7918502807617187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,1,4,balanced,2.401685791015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,1,4,power_law_1.01,0.27720191955566403
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,4,1,power_law_1.2,5.1357751464843755
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,1,4,power_law_1.01,0.2758387184143066
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,1,4,balanced,3.552115173339844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,1,4,power_law_1.01,0.5349734497070313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,1,4,power_law_1.01,0.5610623931884765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,1,4,power_law_1.01,0.5756787109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,1,4,power_law_1.01,0.5808768081665039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,1,4,balanced,15.400037841796877
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,1,4,power_law_1.01,0.5957081604003907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,1,4,power_law_1.01,0.613894386291504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,1,4,power_law_1.01,0.6401522827148438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,1,4,power_law_1.01,0.6462566375732421
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,1,4,power_law_1.01,0.6470617675781251
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,1,4,power_law_1.01,0.6738944244384766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,1,4,balanced,4.736246948242187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,1,4,power_law_1.01,0.7409855651855468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,1,4,power_law_1.01,0.8262834930419922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,1,4,power_law_1.01,0.987097625732422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,4,1,power_law_1.2,2.2446380615234376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,1,4,power_law_1.01,1.2524505615234376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,1,4,power_law_1.01,1.6575923156738281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,1,4,power_law_1.01,2.2579890441894532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,1,4,power_law_1.01,2.6103808593749998
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,1,4,power_law_1.01,3.9778778076171877
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,1,4,balanced,21.96433349609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,1,4,power_law_1.01,4.7879052734375005
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,1,4,power_law_1.01,0.12186112403869628
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,1,4,power_law_1.01,0.1629542350769043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,1,4,power_law_1.01,0.2031065559387207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,1,4,power_law_1.01,0.21331455230712892
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,1,4,power_law_1.01,0.21446271896362307
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,1,4,power_law_1.01,0.2163315200805664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,1,4,power_law_1.01,0.2182169532775879
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,1,4,power_law_1.01,0.22028928756713867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,1,4,power_law_1.01,0.2225881576538086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,1,4,power_law_1.01,0.2244927978515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,1,4,power_law_1.01,8.252247314453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,1,4,power_law_1.01,0.24967296600341796
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,1,4,power_law_1.01,0.25893119812011717
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,1,4,power_law_1.01,0.22884479522705078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,1,4,power_law_1.01,0.26380544662475586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,1,4,power_law_1.01,0.3006617546081543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,1,4,power_law_1.01,0.34712192535400394
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,1,4,power_law_1.01,0.409436149597168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,1,4,power_law_1.01,0.540579833984375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,1,4,power_law_1.01,0.9299609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,1,4,power_law_1.01,9.621683349609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,1,4,power_law_1.01,0.7435993957519531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,4,1,power_law_1.2,2.9680914306640624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,1,4,power_law_1.01,15.123171386718749
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,1,4,power_law_1.01,1.4035084533691407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,1,4,power_law_1.01,2.085379791259766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,1,4,power_law_1.01,2.5396237182617187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,1,4,power_law_1.2,0.26428287506103515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,1,4,power_law_1.2,0.27197952270507814
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,1,4,power_law_1.2,0.5248012924194336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,1,4,power_law_1.2,0.5610329437255859
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,1,4,power_law_1.2,0.5303104019165039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,1,4,power_law_1.2,0.5827174377441406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,1,4,power_law_1.2,0.5984588623046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,1,4,power_law_1.2,0.565722885131836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,1,4,power_law_1.2,0.6346432113647461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,1,4,power_law_1.2,0.6504370880126953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,1,4,power_law_1.01,3.670271911621094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,1,4,power_law_1.2,0.6603046417236328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,1,4,power_law_1.2,0.6765811157226562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,1,4,power_law_1.01,5.7915380859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,1,4,power_law_1.2,0.7493196868896485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,1,4,power_law_1.2,0.8197875213623046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,1,4,power_law_1.2,1.0929599761962892
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,1,4,power_law_1.01,7.986735229492187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,1,4,power_law_1.2,1.330127410888672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,1,4,power_law_1.2,1.5055526733398437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,1,4,power_law_1.2,2.4248115539550783
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,1,4,power_law_1.01,27.200283203125004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,1,4,power_law_1.2,3.2788159179687497
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,1,4,power_law_1.2,4.004618225097657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,1,4,power_law_1.01,34.6218310546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,1,4,power_law_1.01,10.640234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,1,4,power_law_1.2,6.843533935546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,1,4,power_law_1.2,0.12167807579040528
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,1,4,power_law_1.2,0.14979711532592774
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,1,4,power_law_1.2,0.20347007751464843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,1,4,power_law_1.2,0.2134105682373047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,1,4,power_law_1.2,0.21444608688354494
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,1,4,power_law_1.2,0.21635200500488283
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,1,4,power_law_1.2,0.21864448547363283
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,1,4,power_law_1.2,0.22085248947143557
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,1,4,power_law_1.2,0.22327680587768556
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,1,4,power_law_1.2,0.22480512619018556
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,1,4,power_law_1.2,0.2298918342590332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,1,4,power_law_1.2,0.24471807479858398
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,1,4,power_law_1.2,0.2540032005310059
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,1,4,power_law_1.2,8.068257446289063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,1,4,power_law_1.2,0.265163516998291
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,1,4,power_law_1.2,0.30332672119140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,1,4,power_law_1.2,0.35040382385253904
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,1,4,power_law_1.2,10.430703124999999
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,1,4,power_law_1.2,0.3803007888793945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,1,4,power_law_1.2,0.5478617477416992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,1,4,power_law_1.2,0.9502528381347656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,1,4,power_law_1.2,0.6745203399658204
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,1,4,power_law_1.2,1.2255974578857423
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,1,4,power_law_1.2,2.5143641662597656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,1,4,power_law_1.2,1.831480255126953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,1,4,power_law_1.2,20.996611328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,1,4,power_law_1.2,3.5003430175781247
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,2,2,balanced,0.13228799819946288
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,2,2,balanced,0.1815283203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,2,2,balanced,0.36204288482666014
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,2,2,balanced,0.4727936172485352
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,2,2,balanced,0.47669761657714843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,2,2,balanced,0.47966846466064456
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,2,2,balanced,0.4879526519775391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,2,2,balanced,0.49093887329101565
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,2,2,balanced,0.4996147155761719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,2,2,balanced,0.5017369461059571
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,2,2,balanced,0.5111347198486328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,2,2,balanced,0.5219148635864258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,2,2,balanced,0.5324057769775391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,2,2,balanced,0.548908805847168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,2,2,balanced,0.7334323120117188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,2,2,balanced,0.6859033966064454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,2,2,balanced,0.7337420654296876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,2,2,balanced,1.0837273406982422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,2,2,balanced,1.303014373779297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,1,4,power_law_1.2,4.841204528808594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,1,4,power_law_1.01,52.233432617187496
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,2,2,balanced,1.9248384094238282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,1,4,power_law_1.2,29.6005126953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,2,2,balanced,2.3395277404785153
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,2,2,balanced,3.599862976074219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,2,2,balanced,4.740753784179687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,1,4,power_law_1.2,8.385115966796874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,2,2,balanced,0.07728896141052247
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,2,2,balanced,0.10354944229125977
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,2,2,balanced,0.17318527221679686
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,2,2,balanced,6.902063598632813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,2,2,balanced,0.1826304054260254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,2,2,balanced,0.1866854476928711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,2,2,balanced,0.1850009536743164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,2,2,balanced,0.1870297622680664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,2,2,balanced,0.18623359680175783
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,2,2,balanced,0.18899839401245117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,2,2,balanced,0.18957183837890626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,2,2,balanced,0.19513471603393556
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,2,2,balanced,0.1966604804992676
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,2,2,balanced,0.2029350471496582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,2,2,balanced,0.21021055221557616
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,2,2,balanced,0.21842432022094727
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,2,2,balanced,0.2255615997314453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,2,2,balanced,0.22876800537109374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,2,2,balanced,0.316628475189209
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,2,2,balanced,0.3500697708129883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,2,2,balanced,0.4744960021972656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,2,2,balanced,9.302540893554688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,1,4,power_law_1.2,36.936728515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,2,2,balanced,0.6069337463378905
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,2,2,balanced,0.9147622680664063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,2,2,balanced,1.2943705749511718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,2,2,balanced,2.0098841857910155
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,2,2,power_law_1.01,0.15942015647888183
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,2,2,balanced,2.513469390869141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,2,2,power_law_1.01,0.1770124816894531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,2,2,power_law_1.01,0.23242879867553712
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,2,2,power_law_1.01,0.4553049468994141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,2,2,power_law_1.01,0.4787814331054688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,1,4,power_law_1.2,50.715190429687496
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,2,2,balanced,3.687682495117188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,2,2,power_law_1.01,0.48912384033203127
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,2,2,power_law_1.01,0.4997555160522461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,2,2,power_law_1.01,0.5150860977172852
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,2,2,power_law_1.01,0.5452851104736328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,2,2,power_law_1.01,0.5701273727416992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,2,2,power_law_1.01,0.5846054458618164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,2,2,balanced,4.949899597167969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,2,2,power_law_1.01,0.583009262084961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,2,2,power_law_1.01,0.6380505752563477
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,2,2,power_law_1.01,0.7027008056640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,2,2,balanced,15.025603027343749
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,2,2,power_law_1.01,0.7888883209228515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,2,2,power_law_1.01,0.8911424255371093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,2,2,power_law_1.01,1.0886540985107422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,2,2,power_law_1.01,1.476486358642578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,2,2,power_law_1.01,1.8595660400390623
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,2,2,power_law_1.01,2.7767041015625002
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,2,2,power_law_1.01,2.8514727783203124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,2,2,power_law_1.01,4.972901000976562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,2,2,power_law_1.01,0.08939135551452637
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,2,2,power_law_1.01,0.10431872367858888
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,2,2,balanced,21.15301513671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,2,2,power_law_1.01,0.11844736099243165
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,2,2,power_law_1.01,0.1778598403930664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,2,2,power_law_1.01,0.18339071273803711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,2,2,power_law_1.01,0.18570880889892577
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,2,2,power_law_1.01,0.18570240020751955
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,2,2,power_law_1.01,0.18852352142333983
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,2,2,power_law_1.01,0.18951040267944336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,2,2,power_law_1.01,6.44806640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,2,2,power_law_1.01,0.19023872375488282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,2,2,power_law_1.01,0.19483264923095703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,2,2,power_law_1.01,0.22451456069946288
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,2,2,power_law_1.01,0.2249817657470703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,2,2,power_law_1.01,0.2689715194702148
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,2,2,power_law_1.01,0.29183231353759764
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,2,2,power_law_1.01,0.20980735778808596
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,2,2,power_law_1.01,0.31113088607788086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,2,2,power_law_1.01,0.384637451171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,2,2,power_law_1.01,0.4941632080078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,2,2,power_law_1.01,0.6898355102539062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,2,2,power_law_1.01,11.040029296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,2,2,power_law_1.01,0.9544409942626952
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,2,2,power_law_1.01,1.6253990173339843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,2,2,power_law_1.01,2.861505432128906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,2,2,power_law_1.01,1.3506150817871094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,2,2,power_law_1.2,0.15287808418273924
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,2,2,power_law_1.2,0.1851353645324707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,2,2,power_law_1.2,0.2663808059692383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,2,2,power_law_1.2,0.4548646545410156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,2,2,power_law_1.01,3.4463565063476564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,2,2,power_law_1.2,0.45751808166503904
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,2,2,power_law_1.2,0.4966400146484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,2,2,power_law_1.2,0.5002956771850586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,2,2,power_law_1.2,0.4912934494018555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,2,2,power_law_1.2,0.5482137680053711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,2,2,power_law_1.2,0.5808000183105468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,2,2,power_law_1.2,0.5832691192626953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,2,2,power_law_1.2,0.6013631820678711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,2,2,power_law_1.2,0.6279167938232422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,2,2,power_law_1.2,0.7372108459472656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,2,2,power_law_1.2,0.8164249420166015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,2,2,power_law_1.2,0.929117431640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,2,2,power_law_1.2,1.2060671997070311
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,2,2,power_law_1.01,15.433192138671876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,1,4,power_law_1.2,11.85771240234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,2,2,power_law_1.2,1.4894873046875001
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,2,2,power_law_1.2,1.9211878967285156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,2,2,power_law_1.01,5.19239013671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,2,2,power_law_1.01,6.40265869140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,2,2,power_law_1.2,2.89517578125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,2,2,power_law_1.2,3.6208743286132816
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,2,2,power_law_1.01,22.22170166015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,2,2,power_law_1.2,0.08860671997070313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,2,2,power_law_1.2,0.10406784057617187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,2,2,power_law_1.2,0.11865856170654296
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,2,2,power_law_1.2,4.876080627441406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,2,2,power_law_1.2,0.1817420768737793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,2,2,power_law_1.2,6.565618896484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,2,2,power_law_1.2,0.17905023574829101
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,2,2,power_law_1.2,0.18417024612426758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,2,2,power_law_1.2,0.1879884719848633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,2,2,power_law_1.2,0.1876095962524414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,2,2,power_law_1.2,0.18898944854736327
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,2,2,power_law_1.2,0.1922675132751465
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,2,2,power_law_1.2,0.1940889549255371
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,2,2,power_law_1.2,0.20925184249877932
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,2,2,power_law_1.2,0.2183091163635254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,2,2,power_law_1.2,0.2322534370422363
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,2,2,power_law_1.2,0.2742655944824219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,2,2,power_law_1.2,0.28766080856323245
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,2,2,power_law_1.2,0.30473855972290037
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,2,2,power_law_1.2,0.39121921539306637
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,2,2,power_law_1.2,9.392679443359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,2,2,power_law_1.2,0.4945049667358399
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,2,2,power_law_1.2,0.6617996978759766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,2,2,power_law_1.2,0.8763302612304689
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,2,2,power_law_1.2,1.7731111145019534
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,2,2,power_law_1.2,14.565450439453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,2,2,power_law_1.2,1.3676864624023437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,4,1,balanced,0.09719615936279297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,2,2,power_law_1.2,2.5079718017578125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,4,1,balanced,0.3118144035339355
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,4,1,balanced,0.17651519775390626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,2,2,power_law_1.01,29.34896240234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,2,2,power_law_1.2,3.8640524291992184
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,4,1,balanced,0.30160640716552733
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,4,1,balanced,0.31245567321777346
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,4,1,balanced,0.3185113525390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,4,1,balanced,0.3155283164978028
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,4,1,balanced,0.3050464057922363
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,4,1,balanced,0.31814464569091794
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,4,1,balanced,0.32094974517822267
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,4,1,balanced,0.324021110534668
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,4,1,balanced,0.3571891021728516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,4,1,balanced,0.3610483169555664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,4,1,balanced,0.3685523223876953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,4,1,balanced,0.3736383819580078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,4,1,balanced,0.5143008041381836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,4,1,balanced,0.5321785736083984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,2,2,power_law_1.2,5.216931762695312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,4,1,balanced,0.8741593933105468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,4,1,balanced,0.31806976318359376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,4,1,balanced,2.185051574707031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,4,1,balanced,0.3139904022216797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,4,1,balanced,3.2299597167968748
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,2,2,power_law_1.2,23.067326660156247
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,4,1,balanced,4.668149108886719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,4,1,balanced,1.0831353759765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,4,1,balanced,1.6640953063964843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,4,1,balanced,6.28876220703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,4,1,balanced,0.06189311981201172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,4,1,balanced,0.17896511077880858
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,4,1,balanced,0.17992767333984375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,4,1,balanced,0.18121408462524413
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,4,1,balanced,0.1821664047241211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,4,1,balanced,0.0993177604675293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,4,1,balanced,8.8189599609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,4,1,balanced,0.19048128128051758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,4,1,balanced,0.19220735549926757
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,4,1,balanced,0.17956928253173826
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,4,1,balanced,0.2120479965209961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,4,1,balanced,0.21848640441894532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,4,1,balanced,0.18358272552490235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,4,1,balanced,0.18166080474853516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,4,1,balanced,0.3011916732788086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,4,1,balanced,0.1827961540222168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,4,1,balanced,0.4486387252807617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,4,1,balanced,0.5714796829223633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,4,1,balanced,0.18563264846801758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,4,1,balanced,0.9232479858398438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,4,1,balanced,1.20109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,4,1,balanced,1.8071328735351564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,4,1,balanced,2.348926696777344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,4,1,power_law_1.01,0.09603327751159668
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,4,1,balanced,3.5714413452148435
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,4,1,balanced,0.20732608795166016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,4,1,balanced,4.750950317382812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,4,1,balanced,0.22372800827026368
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,4,1,balanced,0.2890649604797363
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,4,1,power_law_1.01,0.20285951614379885
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,4,1,power_law_1.01,0.16567359924316408
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,4,1,power_law_1.01,0.35928958892822266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,4,1,power_law_1.01,0.3215711975097656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,2,2,power_law_1.2,7.479138793945313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,4,1,power_law_1.01,0.3049446487426758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,4,1,power_law_1.01,0.35633728027343753
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,4,1,power_law_1.01,0.30184255599975585
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,4,1,power_law_1.01,0.37249279022216797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,4,1,power_law_1.01,0.4027577590942383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,4,1,power_law_1.01,0.39668033599853514
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,4,1,power_law_1.01,0.3969472122192383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,4,1,power_law_1.01,0.32575168609619143
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,4,1,power_law_1.01,0.42236480712890623
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,4,1,power_law_1.01,0.6267462539672851
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,4,1,power_law_1.01,0.7402156829833985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,4,1,power_law_1.01,1.0338553619384765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,2,2,power_law_1.2,33.301249999999996
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,4,1,power_law_1.01,0.35322753906250004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,4,1,power_law_1.01,0.31464576721191406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,4,1,power_law_1.01,1.7363058471679687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,4,1,power_law_1.01,2.3156390380859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,4,1,power_law_1.01,0.09827584266662598
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,4,1,power_law_1.01,0.4702880096435547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,4,1,power_law_1.01,0.06188992023468017
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,4,1,power_law_1.01,3.3790164184570317
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,4,1,power_law_1.01,0.18188352584838868
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,4,1,power_law_1.01,1.2143270111083984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,4,1,power_law_1.01,0.18106367111206054
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,4,1,power_law_1.01,0.181408634185791
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,4,1,power_law_1.01,0.1798246383666992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,4,1,power_law_1.01,0.1884409523010254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,4,1,power_law_1.01,0.19565120697021485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,4,1,power_law_1.01,0.19430463790893554
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,4,1,power_law_1.01,0.2040166473388672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,4,1,power_law_1.01,0.22700159072875978
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,4,1,power_law_1.01,0.22909376144409183
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,4,1,power_law_1.01,0.2442348861694336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,4,1,power_law_1.01,4.357499389648437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,4,1,power_law_1.01,0.26452991485595706
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,4,1,power_law_1.01,0.34014785766601563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,4,1,power_law_1.01,0.389633903503418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,4,1,power_law_1.01,0.17908544540405275
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,4,1,power_law_1.01,0.12431039810180664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,4,1,power_law_1.01,0.5226604843139648
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,4,1,power_law_1.01,0.6242425537109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,4,1,power_law_1.01,0.17997312545776367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,4,1,power_law_1.01,0.9187071990966796
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,4,1,power_law_1.01,1.2590995025634766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,4,1,power_law_1.2,0.1660812759399414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,4,1,power_law_1.2,0.18380544662475587
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,4,1,power_law_1.2,0.3083097648620606
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,4,1,power_law_1.01,6.45319580078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,4,1,power_law_1.01,1.899040069580078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,4,1,power_law_1.2,0.09433856010437011
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,4,1,power_law_1.2,0.28993343353271483
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,4,1,power_law_1.01,2.4304928588867187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,4,1,power_law_1.2,0.3660947036743164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,4,1,power_law_1.2,0.3472275161743164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,4,1,power_law_1.2,0.32505470275878906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,4,1,power_law_1.2,0.36077247619628905
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,4,1,power_law_1.2,0.36452991485595704
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,4,1,power_law_1.2,0.40034046173095705
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,4,1,power_law_1.2,0.39726016998291014
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,4,1,power_law_1.2,0.39888256072998046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,4,1,power_law_1.2,0.4384652709960937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,4,1,power_law_1.2,0.4926655960083008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,4,1,power_law_1.2,0.3516320037841797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,4,1,power_law_1.2,0.5933203125000001
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,4,1,power_law_1.2,0.6775161743164062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,4,1,power_law_1.01,8.581865234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,4,1,power_law_1.2,0.9596697235107422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,4,1,power_law_1.2,0.3127699279785156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,4,1,power_law_1.01,3.6414599609375005
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,4,1,power_law_1.2,1.7864927673339843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,4,1,power_law_1.2,2.273645477294922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,4,1,power_law_1.2,0.06250239849090576
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,4,1,power_law_1.01,4.810968322753906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,4,1,power_law_1.2,0.18928064346313475
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,4,1,power_law_1.2,1.223114242553711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,4,1,power_law_1.2,0.18268352508544922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,4,1,power_law_1.2,0.18131839752197265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,4,1,power_law_1.2,0.17869760513305663
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,4,1,power_law_1.2,0.1828223991394043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,4,1,power_law_1.2,0.1863225555419922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,4,1,power_law_1.2,0.19510143280029296
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,4,1,power_law_1.2,0.19676031112670897
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,4,1,power_law_1.2,0.20139328002929688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,4,1,power_law_1.2,3.358184204101563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,4,1,power_law_1.2,0.22703615188598633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,4,1,power_law_1.2,0.0932857608795166
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,4,1,power_law_1.2,0.25216896057128907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,4,1,power_law_1.2,0.27505279541015626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,4,1,power_law_1.2,0.33065345764160153
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,4,1,power_law_1.2,0.17405311584472655
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,4,1,power_law_1.2,0.37948352813720704
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,4,1,power_law_1.2,0.11173760414123535
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,4,1,power_law_1.2,0.5251878356933594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,4,1,power_law_1.2,0.6379359817504883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,4,1,power_law_1.2,4.293348999023438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,4,1,power_law_1.2,0.929134750366211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,4,1,power_law_1.2,0.22980224609375002
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,1,4,balanced,0.07989503860473632
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,1,4,balanced,0.10915712356567382
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,1,4,balanced,0.2015987205505371
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,1,4,balanced,0.4207436752319336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,1,4,balanced,0.7051686096191406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,4,1,power_law_1.2,1.2342745971679687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,1,4,balanced,1.3312371826171874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,1,4,balanced,1.333745880126953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,1,4,balanced,1.33968505859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,1,4,balanced,2.0458277893066406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,1,4,balanced,2.0477005004882813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,1,4,balanced,2.0437606811523437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,1,4,balanced,2.0639974975585935
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,4,1,power_law_1.2,1.8740031433105468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,1,4,balanced,2.067923126220703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,1,4,balanced,2.08513916015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,1,4,balanced,2.1288485717773438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,4,1,power_law_1.2,6.527876586914063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,1,4,balanced,2.139395904541016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,1,4,balanced,2.183221740722656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,1,4,balanced,2.298065948486328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,1,4,balanced,2.4140992736816407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,1,4,balanced,2.6004479980468753
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,1,4,balanced,2.7552972412109376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,1,4,balanced,3.211032409667969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,1,4,balanced,3.4798388671875005
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,4,1,power_law_1.2,2.4728338623046873
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,1,4,balanced,0.05542399883270264
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,1,4,balanced,0.07179264068603516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,1,4,balanced,0.11866239547729492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,1,4,balanced,4.985985412597656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,1,4,balanced,0.35582847595214845
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,1,4,balanced,0.2041753578186035
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,1,4,balanced,0.6801945495605468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,1,4,balanced,6.310933837890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,1,4,balanced,0.6790873718261718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,1,4,balanced,0.6852057647705079
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,1,4,balanced,0.7250176239013671
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,1,4,balanced,0.7407552337646484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,1,4,balanced,0.7313906860351562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,1,4,balanced,0.7434457397460937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,1,4,balanced,0.7365593719482422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,1,4,balanced,0.7497714996337891
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,1,4,balanced,0.7621401977539063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,1,4,balanced,9.868292846679688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,1,4,balanced,0.7654399871826172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,4,1,power_law_1.2,8.548156127929687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,1,4,balanced,0.7904140472412109
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,1,4,balanced,0.8464614105224608
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,1,4,balanced,0.8677926635742187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,1,4,balanced,0.9403507232666015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,1,4,balanced,1.065648651123047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,1,4,balanced,1.1467635345458984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,1,4,balanced,1.3210546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,1,4,power_law_1.01,0.20608768463134766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,1,4,balanced,1.923795166015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,1,4,balanced,2.200853729248047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,1,4,balanced,12.623448486328126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,1,4,power_law_1.01,0.4243724822998047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,1,4,power_law_1.01,0.4989990234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,4,1,power_law_1.2,3.582441101074219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,1,4,power_law_1.01,0.6350604629516601
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,1,4,power_law_1.01,0.7155750274658204
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,1,4,power_law_1.01,0.8210111999511719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,1,4,balanced,3.181481018066406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,1,4,power_law_1.01,1.1091673278808594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,1,4,power_law_1.01,1.1221196746826172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,1,4,balanced,4.000750122070313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,1,4,power_law_1.01,1.6527104187011719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,1,4,power_law_1.01,1.6254911804199217
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,1,4,power_law_1.01,1.6864448547363282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,1,4,power_law_1.01,1.9347570800781249
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,1,4,power_law_1.01,1.9266085815429688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,1,4,power_law_1.01,2.0082252502441404
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,1,4,power_law_1.01,1.992216339111328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,1,4,power_law_1.01,2.005902099609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,1,4,power_law_1.01,2.075792694091797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,1,4,power_law_1.01,2.2692633056640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,1,4,power_law_1.01,2.44163330078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,1,4,power_law_1.01,2.8722738647460937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,1,4,power_law_1.01,3.0326116943359374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,1,4,power_law_1.01,0.11786368370056152
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,1,4,power_law_1.01,0.20269056320190432
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,1,4,power_law_1.01,3.8187982177734376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,1,4,power_law_1.01,0.2606847953796387
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,1,4,power_law_1.01,0.32994049072265624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,1,4,power_law_1.01,4.6100927734375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,1,4,power_law_1.01,0.41705215454101563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,1,4,power_law_1.01,0.4838899230957031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,1,4,power_law_1.01,0.5657126235961913
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,1,4,power_law_1.01,0.5916236877441406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,1,4,power_law_1.01,0.6294105529785157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,1,4,power_law_1.01,6.527194824218751
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,1,4,power_law_1.01,0.6569420623779296
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,1,4,power_law_1.01,0.6898841857910156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,1,4,power_law_1.01,0.6813875579833985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,1,4,power_law_1.01,0.7050994873046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,1,4,power_law_1.01,8.246760864257812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,1,4,power_law_1.01,0.704005126953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,1,4,power_law_1.01,0.7031807708740234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,1,4,power_law_1.01,0.7053644561767578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,1,4,power_law_1.01,0.8066764831542969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,1,4,power_law_1.01,0.8376844787597657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,1,4,power_law_1.01,0.9741158294677735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,1,4,power_law_1.01,1.0254144287109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,1,4,power_law_1.01,1.4039219665527345
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,1,4,power_law_1.01,1.735467529296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,1,4,power_law_1.01,2.1655795288085935
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,1,4,power_law_1.2,0.20718719482421877
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,1,4,power_law_1.01,0.6058623886108399
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,1,4,power_law_1.01,2.7249099731445314
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,1,4,power_law_1.2,0.42647422790527345
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,1,4,power_law_1.2,0.44753150939941405
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,1,4,power_law_1.01,11.896844482421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,1,4,power_law_1.2,0.5409215927124024
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,1,4,power_law_1.2,0.6749887847900391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,1,4,power_law_1.01,3.7117108154296874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,1,4,power_law_1.2,1.0664793395996095
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,1,4,power_law_1.2,1.0661158752441406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,4,1,power_law_1.2,4.824159240722656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,1,4,power_law_1.2,1.6338137817382812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,1,4,power_law_1.01,4.808295593261719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,1,4,power_law_1.2,1.7140287780761718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,1,4,power_law_1.01,15.682532958984376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,1,4,power_law_1.2,1.7074649047851562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,1,4,power_law_1.2,1.8226162719726564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,1,4,power_law_1.2,1.9640269470214844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,1,4,power_law_1.2,0.7996800231933594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,1,4,power_law_1.2,1.9952423095703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,1,4,power_law_1.2,1.9440626525878906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,1,4,power_law_1.2,2.029564208984375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,1,4,power_law_1.2,2.0936793518066406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,1,4,power_law_1.2,2.4384806823730467
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,1,4,power_law_1.2,2.6688385009765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,1,4,power_law_1.2,2.989470825195313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,1,4,power_law_1.2,3.314269409179688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,1,4,power_law_1.2,0.11822208404541015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,1,4,power_law_1.2,4.435979614257812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,1,4,power_law_1.2,0.19983360290527344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,1,4,power_law_1.2,0.21722623825073245
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,1,4,power_law_1.2,0.27398143768310546
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,1,4,power_law_1.2,5.010015869140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,1,4,power_law_1.2,0.3462451171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,1,4,power_law_1.2,0.48706432342529293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,1,4,power_law_1.2,0.6017523193359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,1,4,power_law_1.2,0.5484441757202149
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,1,4,power_law_1.2,0.6208000183105469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,1,4,power_law_1.2,0.6125196838378907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,1,4,power_law_1.2,0.6144550323486329
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,1,4,power_law_1.2,0.6384614562988281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,1,4,power_law_1.2,0.6986470031738281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,1,4,power_law_1.2,0.6975910186767578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,1,4,power_law_1.2,0.6685670471191406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,1,4,power_law_1.2,0.6843186950683593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,1,4,power_law_1.2,0.7461542510986329
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,1,4,power_law_1.2,0.8001113891601562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,1,4,power_law_1.2,7.168914184570312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,1,4,power_law_1.2,0.8402521514892578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,1,4,power_law_1.2,0.9993920135498048
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,1,4,power_law_1.2,8.553578491210938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,1,4,power_law_1.2,1.1103679656982421
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,1,4,power_law_1.2,1.4173146057128907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,1,4,power_law_1.2,1.7338546752929687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,2,2,balanced,0.08035072326660156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,2,2,balanced,0.19400064468383788
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,1,4,power_law_1.2,2.2371725463867187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,2,2,balanced,0.40821247100830077
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,2,2,balanced,0.6963200378417969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,1,4,power_law_1.2,2.9021093749999998
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,2,2,balanced,0.11780351638793944
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,2,2,balanced,1.3292466735839843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,2,2,balanced,1.3303424072265624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,2,2,balanced,1.3345228576660157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,2,2,balanced,1.3406387329101563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,1,4,power_law_1.2,4.454419250488281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,2,2,balanced,1.3500901794433593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,1,4,power_law_1.2,13.608131103515626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,2,2,balanced,1.3543128967285156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,1,4,power_law_1.2,5.49391845703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,2,2,balanced,2.0077810668945313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,2,2,balanced,2.022044219970703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,2,2,balanced,2.0395724487304685
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,2,2,balanced,2.075029754638672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,2,2,balanced,2.0880895996093747
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,2,2,balanced,2.122195129394531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,2,2,balanced,2.227467498779297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,1,4,power_law_1.2,16.147687988281252
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,2,2,balanced,2.354358978271484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,2,2,balanced,2.5651455688476563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,2,2,balanced,0.05268735885620117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,2,2,balanced,2.730858154296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,2,2,balanced,0.07158016204833985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,2,2,balanced,0.11999615669250488
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,2,2,balanced,0.20341375350952146
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,2,2,balanced,3.2066329956054687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,2,2,balanced,0.35548545837402346
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,2,2,balanced,3.452101135253906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,2,2,balanced,0.6782374572753906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,2,2,balanced,0.6811673736572266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,2,2,balanced,0.6812326049804687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,2,2,balanced,0.6885401916503906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,2,2,balanced,0.6911923217773437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,2,2,balanced,0.6975244903564454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,2,2,balanced,0.7524275207519532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,2,2,balanced,4.882574157714844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,2,2,balanced,0.7497894287109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,2,2,balanced,0.7677414703369141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,2,2,balanced,0.7865190124511718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,2,2,balanced,0.7822720336914062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,2,2,balanced,0.7951129913330078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,2,2,balanced,0.8166464233398438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,2,2,balanced,0.8605593872070312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,2,2,balanced,6.173134155273438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,2,2,balanced,0.9341004943847656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,2,2,balanced,0.9940019226074218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,2,2,balanced,1.135304946899414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,2,2,balanced,1.2510208129882812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,2,2,power_law_1.01,0.11781760215759278
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,2,2,power_law_1.01,0.18645759582519533
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,2,2,power_law_1.01,0.31630720138549806
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,2,2,balanced,1.9062950134277343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,2,2,balanced,2.1660992431640627
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,2,2,power_law_1.01,0.47032833099365234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,2,2,power_law_1.01,0.6547468566894532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,2,2,power_law_1.01,0.7754176330566407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,2,2,balanced,9.226795654296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,2,2,balanced,3.176312255859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,2,2,power_law_1.01,1.08587646484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,2,2,power_law_1.01,1.0613184356689453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,2,2,power_law_1.01,1.154224624633789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,2,2,power_law_1.01,1.1259532928466798
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,2,2,balanced,3.9967846679687495
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,2,2,power_law_1.01,1.1525721740722656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,2,2,power_law_1.01,1.7639820861816407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,2,2,balanced,11.74973876953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,2,2,power_law_1.01,1.8482330322265625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,2,2,power_law_1.01,1.8803546142578125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,2,2,power_law_1.01,1.8812403869628906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,2,2,power_law_1.01,1.8977049255371092
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,2,2,power_law_1.01,2.0076211547851566
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,2,2,power_law_1.01,2.222725067138672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,2,2,power_law_1.01,2.2839474487304687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,2,2,power_law_1.01,2.6941375732421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,2,2,power_law_1.01,2.9656216430664064
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,2,2,power_law_1.01,0.07245696067810059
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,2,2,power_law_1.01,0.12009216308593751
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,2,2,power_law_1.01,0.17520767211914062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,2,2,power_law_1.01,3.611971740722656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,2,2,power_law_1.01,0.24731775283813479
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,2,2,power_law_1.01,0.33131649017333986
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,2,2,power_law_1.01,0.4082316970825195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,2,2,power_law_1.01,4.1860107421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,2,2,power_law_1.01,0.5451020812988281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,2,2,power_law_1.01,0.5796428680419922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,2,2,power_law_1.01,0.5560627365112305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,2,2,power_law_1.01,0.5774220657348633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,2,2,power_law_1.01,0.5991321563720703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,2,2,power_law_1.01,0.6497625732421876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,2,2,power_law_1.01,0.6601971435546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,2,2,power_law_1.01,5.566154174804687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,2,2,power_law_1.01,0.6881088256835938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,2,2,power_law_1.01,0.7053734588623046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,2,2,power_law_1.01,0.6999603271484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,2,2,power_law_1.01,0.7226509094238281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,2,2,power_law_1.01,0.7839807891845704
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,2,2,power_law_1.01,0.8237260437011719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,2,2,power_law_1.01,0.9072831726074219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,2,2,power_law_1.01,1.0323661041259764
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,2,2,power_law_1.01,7.178036499023437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,2,2,power_law_1.01,1.2954547119140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,2,2,power_law_1.2,0.11633919715881347
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,2,2,power_law_1.01,1.6048959350585938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,2,2,power_law_1.2,0.18518144607543946
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,2,2,power_law_1.2,0.28101119995117185
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,2,2,power_law_1.01,2.1656704711914063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,2,2,power_law_1.2,0.45293952941894533
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,2,2,power_law_1.01,2.5102029418945313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,2,2,power_law_1.2,0.6129740905761719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,2,2,power_law_1.2,0.7461478424072265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,2,2,power_law_1.2,1.0041702270507813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,2,2,power_law_1.01,10.080408325195313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,2,2,power_law_1.2,1.0894131469726562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,2,2,power_law_1.01,3.5340390014648433
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,2,2,power_law_1.2,1.1603826904296874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,2,2,power_law_1.2,1.1799462127685547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,2,2,power_law_1.2,1.0819929504394532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,2,2,power_law_1.01,4.683627624511719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,2,2,power_law_1.2,1.9265689086914062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,2,2,power_law_1.2,1.8408819580078124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,2,2,power_law_1.01,12.664224853515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,2,2,power_law_1.2,1.892761535644531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,2,2,power_law_1.2,1.895379180908203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,2,2,power_law_1.2,1.9265177917480467
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,2,2,power_law_1.2,2.0348377990722657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,2,2,power_law_1.2,2.2605670166015623
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,2,2,power_law_1.2,2.449157104492188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,2,2,power_law_1.2,2.7168307495117188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,2,2,power_law_1.2,3.0359103393554685
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,2,2,power_law_1.2,0.07222527980804443
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,2,2,power_law_1.2,0.11947135925292969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,2,2,power_law_1.2,0.16591360092163085
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,2,2,power_law_1.2,3.8162942504882813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,2,2,power_law_1.2,0.2260211181640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,2,2,power_law_1.2,0.3086783981323242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,2,2,power_law_1.2,0.3967552185058594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,2,2,power_law_1.2,4.415074462890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,2,2,power_law_1.2,0.5251071929931641
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,2,2,power_law_1.2,0.5556927871704102
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,2,2,power_law_1.2,0.5738956832885742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,2,2,power_law_1.2,0.5919001770019532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,2,2,power_law_1.2,0.5876416015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,2,2,power_law_1.2,0.6478720092773438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,2,2,power_law_1.2,5.698287353515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,2,2,power_law_1.2,0.6594239807128905
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,2,2,power_law_1.2,0.699530258178711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,2,2,power_law_1.2,0.6766976165771484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,2,2,power_law_1.2,0.6940275573730469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,2,2,power_law_1.2,0.746903076171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,2,2,power_law_1.2,0.8118297576904296
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,2,2,power_law_1.2,0.8703475189208983
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,2,2,power_law_1.2,0.994760971069336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,2,2,power_law_1.2,7.467606811523437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,2,2,power_law_1.2,1.0744000244140626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,2,2,power_law_1.2,1.4200755310058593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,2,2,power_law_1.2,1.5920819091796876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,4,1,balanced,0.1190732765197754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,2,2,power_law_1.2,2.291455993652344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,4,1,balanced,0.18821311950683595
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,4,1,balanced,0.08935999870300293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,4,1,balanced,0.42583679199218755
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,2,2,power_law_1.2,2.766921081542969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,4,1,balanced,0.6985190582275391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,2,2,power_law_1.2,10.499322509765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,2,2,power_law_1.2,4.029726867675781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,4,1,balanced,1.3295634460449217
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,4,1,balanced,1.3376698303222656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,4,1,balanced,1.343900146484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,2,2,power_law_1.2,4.919112854003906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,4,1,balanced,1.3614866638183594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,4,1,balanced,1.3552825927734375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,4,1,balanced,1.36195068359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,4,1,balanced,1.3864588928222656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,4,1,balanced,1.3906060791015624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,2,2,power_law_1.2,13.718004150390623
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,4,1,balanced,1.4146739196777343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,4,1,balanced,2.044855041503906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,4,1,balanced,2.060300750732422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,4,1,balanced,2.0946687316894534
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,4,1,balanced,2.178544006347656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,4,1,balanced,2.290063934326172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,4,1,balanced,2.505188446044922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,4,1,balanced,0.06273280143737793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,4,1,balanced,2.7005209350585937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,4,1,balanced,0.08409088134765626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,4,1,balanced,0.12800064086914062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,4,1,balanced,0.2138483238220215
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,4,1,balanced,3.411797180175781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,4,1,balanced,0.385082893371582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,4,1,balanced,0.7373907470703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,4,1,balanced,0.7429452514648437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,4,1,balanced,0.7431769561767578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,4,1,balanced,0.7465440368652343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,4,1,balanced,3.458036499023437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,4,1,balanced,0.7505510711669923
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,4,1,balanced,0.7578233337402344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,4,1,balanced,4.938689270019531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,4,1,balanced,0.7702252960205078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,4,1,balanced,0.7811238098144531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,4,1,balanced,0.874485092163086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,4,1,balanced,0.7649407958984374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,4,1,balanced,0.8837459564208985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,4,1,balanced,0.9364403533935548
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,4,1,balanced,6.315173950195312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,4,1,balanced,0.9208787536621094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,4,1,balanced,0.9498355102539062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,4,1,balanced,0.985296630859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,4,1,balanced,1.0759571075439454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,4,1,balanced,1.2131053161621093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,4,1,balanced,1.339667205810547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,4,1,power_law_1.01,0.08848128318786622
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,4,1,power_law_1.01,0.11831616401672362
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,4,1,power_law_1.01,0.18797887802124022
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,4,1,balanced,2.2929490661621097
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,4,1,power_law_1.01,0.41286464691162106
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,4,1,power_law_1.01,0.5483219146728515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,4,1,balanced,2.0266201782226565
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,4,1,balanced,9.610879516601562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,4,1,power_law_1.01,0.7154041290283203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,4,1,power_law_1.01,0.9879046630859374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,4,1,balanced,3.2529394531249998
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,4,1,power_law_1.01,1.0354476928710938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,4,1,power_law_1.01,1.0586637115478514
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,4,1,power_law_1.01,1.0769849395751954
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,4,1,power_law_1.01,1.1568358612060545
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,4,1,balanced,4.238110046386718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,4,1,power_law_1.01,1.1873107147216797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,4,1,power_law_1.01,1.2275993347167968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,4,1,power_law_1.01,1.2893017578125001
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,4,1,balanced,12.14589111328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,4,1,power_law_1.01,1.8562637329101563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,4,1,power_law_1.01,1.8655059814453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,4,1,power_law_1.01,1.938269500732422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,4,1,power_law_1.01,2.1249542236328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,4,1,power_law_1.01,2.2525541687011716
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,4,1,power_law_1.01,2.552550964355469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,4,1,power_law_1.01,2.8362655639648438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,4,1,power_law_1.01,0.06297599792480468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,4,1,power_law_1.01,0.08471424102783202
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,4,1,power_law_1.01,0.12931520462036133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,4,1,power_law_1.01,3.3782073974609377
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,4,1,power_law_1.01,0.31002368927001955
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,4,1,power_law_1.01,3.921569213867188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,4,1,power_law_1.01,0.3991551971435547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,4,1,power_law_1.01,0.21498559951782226
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,4,1,power_law_1.01,0.5501881790161133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,4,1,power_law_1.01,0.5779814529418945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,4,1,power_law_1.01,0.5914585494995117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,4,1,power_law_1.01,0.6025696182250977
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,4,1,power_law_1.01,0.6325664138793945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,4,1,power_law_1.01,5.27678955078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,4,1,power_law_1.01,0.6534214019775391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,4,1,power_law_1.01,0.6606726074218751
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,4,1,power_law_1.01,0.6729772949218751
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,4,1,power_law_1.01,0.76740478515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,4,1,power_law_1.01,0.7741613006591797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,4,1,power_law_1.01,0.7981843566894531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,4,1,power_law_1.01,0.8596403503417969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,4,1,power_law_1.01,6.885499267578124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,4,1,power_law_1.01,0.9160384368896484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,4,1,power_law_1.01,1.0189548492431642
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,4,1,power_law_1.01,1.1574892425537109
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,4,1,power_law_1.01,1.3600186157226564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,4,1,power_law_1.01,1.6567922973632812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,4,1,power_law_1.2,0.08946175575256347
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,4,1,power_law_1.01,2.1779615783691404
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,4,1,power_law_1.2,0.11808639526367189
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,4,1,power_law_1.01,2.6371166992187502
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,4,1,power_law_1.2,0.19399616241455078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,4,1,power_law_1.2,0.40144897460937495
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,4,1,power_law_1.01,9.626653442382812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,4,1,power_law_1.2,0.499035530090332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,4,1,power_law_1.2,0.6807756805419922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,4,1,power_law_1.01,3.64686279296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,4,1,power_law_1.2,0.9722694396972656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,4,1,power_law_1.2,1.0198310089111327
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,4,1,power_law_1.2,1.0409945678710937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,4,1,power_law_1.2,1.089046401977539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,4,1,power_law_1.2,1.1424691009521486
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,4,1,power_law_1.01,4.609052124023437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,4,1,power_law_1.2,1.1838451385498048
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,4,1,power_law_1.2,1.2526777648925782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,4,1,power_law_1.01,12.178187255859374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,4,1,power_law_1.2,1.3546937561035155
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,4,1,power_law_1.2,1.8652685546874999
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,4,1,power_law_1.2,1.9018829345703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,4,1,power_law_1.2,1.9758055114746091
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,4,1,power_law_1.2,2.1906335449218752
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,4,1,power_law_1.2,2.314076232910156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,4,1,power_law_1.2,2.6497747802734373
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,4,1,power_law_1.2,2.8841088867187503
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,4,1,power_law_1.2,0.06266111850738525
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,4,1,power_law_1.2,0.08558591842651367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,4,1,power_law_1.2,3.4831994628906253
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,4,1,power_law_1.2,0.12928640365600585
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,4,1,power_law_1.2,0.20020288467407227
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,4,1,power_law_1.2,0.28257791519165043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,4,1,power_law_1.2,0.37388160705566403
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,4,1,power_law_1.2,0.5411743927001953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,4,1,power_law_1.2,4.078575439453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,4,1,power_law_1.2,0.5650086212158203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,4,1,power_law_1.2,0.5797663879394531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,4,1,power_law_1.2,0.6049907302856445
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,4,1,power_law_1.2,0.6283103942871093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,4,1,power_law_1.2,0.6465689849853515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,4,1,power_law_1.2,0.6610476684570312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,4,1,power_law_1.2,5.407228393554687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,4,1,power_law_1.2,0.6839904022216797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,4,1,power_law_1.2,0.7827219390869141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,4,1,power_law_1.2,0.8224723052978515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,4,1,power_law_1.2,0.8160659027099608
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,4,1,power_law_1.2,0.8970464324951172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,4,1,power_law_1.2,0.930216293334961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,4,1,power_law_1.2,6.97112548828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,4,1,power_law_1.2,1.079442596435547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,4,1,power_law_1.2,1.197073287963867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,4,1,power_law_1.2,1.42649658203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,1,4,balanced,0.07381631851196288
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,1,4,balanced,0.09792127609252929
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,1,4,balanced,0.15770496368408202
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,4,1,power_law_1.2,1.8029997253417966
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,1,4,balanced,0.34825855255126953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,4,1,power_law_1.2,2.1349197387695313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,1,4,balanced,0.6076006317138672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,1,4,balanced,1.1408806610107423
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,4,1,power_law_1.2,9.572575073242188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,1,4,balanced,1.1395865631103514
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,1,4,balanced,1.145703659057617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,4,1,power_law_1.2,2.783671569824219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,4,1,power_law_1.2,3.6985189819335935
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,1,4,balanced,1.773107147216797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,1,4,balanced,1.7770162963867187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,1,4,balanced,1.7828466796874998
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,1,4,balanced,1.803678741455078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,4,1,power_law_1.2,4.754676513671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,1,4,balanced,1.780755157470703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,1,4,balanced,1.8148927307128908
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,1,4,balanced,1.8486553955078127
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,1,4,balanced,1.8639961242675782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,4,1,power_law_1.2,12.27753662109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,1,4,balanced,1.901661376953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,1,4,balanced,1.9961599731445312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,1,4,balanced,2.0908953857421873
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,1,4,balanced,2.234544677734375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,1,4,balanced,2.3586457824707034
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,1,4,balanced,0.049328641891479494
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,1,4,balanced,0.06475776195526123
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,1,4,balanced,2.7165927124023437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,1,4,balanced,0.10591872215270996
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,1,4,balanced,0.17924095153808595
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,1,4,balanced,2.965273742675781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,1,4,balanced,0.3102579116821289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,1,4,balanced,0.588545265197754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,1,4,balanced,0.5858457565307618
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,1,4,balanced,0.5922547149658203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,1,4,balanced,0.6331750488281249
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,1,4,balanced,4.286725158691406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,1,4,balanced,0.6343910217285156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,1,4,balanced,0.6362380981445312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,1,4,balanced,0.638524169921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,1,4,balanced,0.6408665466308594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,1,4,balanced,0.6512780761718749
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,1,4,balanced,0.6665740966796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,1,4,balanced,0.6838822174072265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,1,4,balanced,5.3968615722656255
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,1,4,balanced,0.7320729827880859
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,1,4,balanced,0.7549273681640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,1,4,balanced,0.6547737884521484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,1,4,balanced,0.8268940734863282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,1,4,balanced,0.9012518310546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,1,4,balanced,1.0031423950195313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,1,4,balanced,1.1688256072998047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,1,4,power_law_1.01,0.15800576210021972
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,1,4,balanced,1.6850790405273437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,1,4,balanced,1.9278028869628905
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,1,4,power_law_1.01,0.34979583740234377
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,1,4,power_law_1.01,0.40810623168945315
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,1,4,balanced,8.42892822265625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,1,4,power_law_1.01,0.5608230209350585
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,1,4,balanced,2.845672912597656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,1,4,power_law_1.01,0.6427584075927735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,1,4,power_law_1.01,0.7682828521728515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,1,4,power_law_1.01,0.9182592010498046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,1,4,power_law_1.01,0.9812838745117187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,1,4,balanced,3.7611315917968753
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,1,4,power_law_1.01,1.3063909912109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,1,4,power_law_1.01,1.4175360107421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,1,4,power_law_1.01,1.4401971435546874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,1,4,power_law_1.01,1.4823628234863282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,1,4,power_law_1.01,1.5804492187499999
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,1,4,balanced,10.871942138671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,1,4,power_law_1.01,1.6378265380859376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,1,4,power_law_1.01,1.7106866455078127
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,1,4,power_law_1.01,1.7335897827148437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,1,4,power_law_1.01,1.8169625854492188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,1,4,power_law_1.01,1.9268608093261719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,1,4,power_law_1.01,2.0546943664550783
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,1,4,power_law_1.01,2.353260803222656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,1,4,power_law_1.01,2.6395956420898434
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,1,4,power_law_1.01,0.10540160179138183
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,1,4,power_law_1.01,0.17923328399658203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,1,4,power_law_1.01,0.2120319938659668
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,1,4,power_law_1.01,3.2686578369140626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,1,4,power_law_1.01,0.28078975677490237
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,1,4,power_law_1.01,0.3347942352294922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,1,4,power_law_1.01,3.8121829223632817
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,1,4,power_law_1.01,0.47755905151367184
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,1,4,power_law_1.01,0.4505395126342774
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,1,4,power_law_1.01,0.4790323257446289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,1,4,power_law_1.01,0.5076467132568359
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,1,4,power_law_1.01,0.5507519912719727
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,1,4,power_law_1.01,0.5458291244506837
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,1,4,power_law_1.01,5.783582763671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,1,4,power_law_1.01,0.5619379043579101
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,1,4,power_law_1.01,0.583691520690918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,1,4,power_law_1.01,0.5974348831176758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,1,4,power_law_1.01,0.6237555313110351
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,1,4,power_law_1.01,0.5985126495361328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,1,4,power_law_1.01,0.6615347290039063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,1,4,power_law_1.01,6.936564331054687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,1,4,power_law_1.01,0.7331187438964843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,1,4,power_law_1.01,0.8183065795898438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,1,4,power_law_1.01,0.9053004455566406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,1,4,power_law_1.01,1.1320384216308592
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,1,4,power_law_1.01,1.4632730102539062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,1,4,power_law_1.2,0.15869312286376952
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,1,4,power_law_1.01,1.9014976501464844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,1,4,power_law_1.01,9.573199462890624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,1,4,power_law_1.2,0.3587071990966797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,1,4,power_law_1.2,0.37684352874755855
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,1,4,power_law_1.01,2.4522189331054687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,1,4,power_law_1.2,0.5167782211303711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,1,4,power_law_1.2,0.6124761581420899
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,1,4,power_law_1.01,0.39590145111083985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,1,4,power_law_1.2,0.7007308959960937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,1,4,power_law_1.01,3.5456588745117186
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,1,4,power_law_1.2,0.8610099029541015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,1,4,power_law_1.2,0.9830502319335939
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,1,4,power_law_1.2,1.3855052185058594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,1,4,power_law_1.2,1.42671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,1,4,power_law_1.2,1.4193548583984374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,1,4,power_law_1.01,4.440919189453124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,1,4,power_law_1.2,1.539156494140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,1,4,power_law_1.2,1.5421133422851563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,1,4,power_law_1.2,1.6215769958496093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,1,4,power_law_1.2,1.7286886596679687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,1,4,power_law_1.2,1.766950378417969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,1,4,power_law_1.2,1.8113804626464844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,1,4,power_law_1.2,2.0679667663574217
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,1,4,power_law_1.2,1.9938252258300782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,1,4,power_law_1.2,2.530128631591797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,1,4,power_law_1.2,2.7383847045898437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,1,4,power_law_1.01,13.977170410156251
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,1,4,power_law_1.2,0.10532352447509766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,1,4,power_law_1.2,0.18423168182373045
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,1,4,power_law_1.2,0.19054208755493166
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,1,4,power_law_1.2,3.5595687866210937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,1,4,power_law_1.2,0.2579494476318359
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,1,4,power_law_1.2,0.3351103973388672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,1,4,power_law_1.2,0.3563251113891602
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,1,4,power_law_1.2,0.5013568115234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,1,4,power_law_1.2,4.436853637695313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,1,4,power_law_1.2,0.48240768432617187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,1,4,power_law_1.2,0.4980812835693359
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,1,4,power_law_1.2,0.5075519943237305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,1,4,power_law_1.2,0.5154803085327149
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,1,4,power_law_1.2,5.534501342773437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,1,4,power_law_1.2,0.5704998397827149
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,1,4,power_law_1.2,0.5744166564941406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,1,4,power_law_1.2,0.5991616058349609
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,1,4,power_law_1.2,0.5486502456665039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,1,4,power_law_1.2,0.5958092880249024
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,1,4,power_law_1.2,0.649533462524414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,1,4,power_law_1.2,0.733465576171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,1,4,power_law_1.2,7.275159301757813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,1,4,power_law_1.2,0.7522112274169922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,1,4,power_law_1.2,0.8679373168945312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,1,4,power_law_1.2,0.9658188629150392
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,1,4,power_law_1.2,1.2238169860839843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,2,2,balanced,0.07336063861846924
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,1,4,power_law_1.2,1.5325106811523437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,2,2,balanced,0.09837568283081055
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,2,2,balanced,0.16505088806152343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,1,4,power_law_1.2,2.0259263610839846
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,2,2,balanced,0.34875518798828126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,2,2,balanced,0.6000294494628906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,1,4,power_law_1.2,2.5929931640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,1,4,power_law_1.2,10.98353271484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,2,2,balanced,1.138001937866211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,2,2,balanced,1.135447006225586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,1,4,power_law_1.2,3.5682290649414066
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,2,2,balanced,1.1404595184326172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,2,2,balanced,1.1464934539794922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,2,2,balanced,1.1571276855468748
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,2,2,balanced,1.1598233795166015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,2,2,balanced,1.7425280761718749
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,2,2,balanced,1.7536408996582034
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,2,2,balanced,1.7673370361328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,2,2,balanced,1.7967564392089845
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,2,2,balanced,1.8153741455078127
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,2,2,balanced,1.8449612426757813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,2,2,balanced,1.9364915466308594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,1,4,power_law_1.2,14.787137451171876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,2,2,balanced,2.0384909057617184
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,2,2,balanced,2.2015155029296873
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,2,2,balanced,2.334714813232422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,2,2,balanced,0.04822400093078613
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,2,2,balanced,0.06545792102813722
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,2,2,balanced,2.716468505859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,2,2,balanced,0.10547327995300293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,1,4,power_law_1.2,4.642946472167969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,2,2,balanced,2.9595828247070313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,2,2,balanced,0.18001792907714845
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,2,2,balanced,0.30815103530883786
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,2,2,balanced,0.5910745620727539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,2,2,balanced,0.5908044815063477
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,2,2,balanced,0.5895999908447266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,2,2,balanced,0.5926464080810547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,2,2,balanced,0.5975206375122071
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,2,2,balanced,0.6032486343383789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,2,2,balanced,0.6476646423339844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,2,2,balanced,4.215635070800781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,2,2,balanced,0.6467021179199219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,2,2,balanced,0.6569446563720703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,2,2,balanced,0.664163818359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,2,2,balanced,0.6756902313232421
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,2,2,balanced,0.6976780700683594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,2,2,balanced,0.7119846343994141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,2,2,balanced,5.287479248046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,2,2,balanced,0.7540761566162109
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,2,2,balanced,0.809850845336914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,2,2,balanced,0.8629337310791015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,2,2,balanced,0.9888909149169922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,2,2,balanced,1.0962393951416014
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,2,2,power_law_1.01,0.0968166446685791
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,2,2,power_law_1.01,0.1571788787841797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,2,2,power_law_1.01,0.2696870422363281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,2,2,balanced,1.7566925048828126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,2,2,power_law_1.01,0.41509246826171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,2,2,power_law_1.01,0.5094271850585937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,2,2,balanced,7.839380493164063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,2,2,power_law_1.01,0.6578880310058594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,2,2,balanced,2.7533108520507814
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,2,2,power_law_1.01,0.8879615783691406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,2,2,balanced,1.9087487792968751
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,2,2,power_law_1.01,0.9211980438232421
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,2,2,power_law_1.01,0.9497023773193358
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,2,2,power_law_1.01,0.9583014678955077
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,2,2,balanced,3.698414001464844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,2,2,power_law_1.01,0.9891276550292968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,2,2,balanced,9.78722412109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,2,2,power_law_1.01,1.4516224670410156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,2,2,power_law_1.01,1.5470271301269531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,2,2,power_law_1.01,1.57291015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,2,2,power_law_1.01,1.5948800659179687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,2,2,power_law_1.01,1.6609432983398438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,2,2,power_law_1.01,1.7119923400878907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,2,2,power_law_1.01,1.8585049438476564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,2,2,power_law_1.01,1.9426661682128905
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,2,2,power_law_1.01,2.2150604248046877
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,2,2,power_law_1.01,0.06484352111816406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,2,2,power_law_1.01,2.502053070068359
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,2,2,power_law_1.01,0.10515711784362794
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,2,2,power_law_1.01,0.15253376007080077
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,2,2,power_law_1.01,3.1007730102539064
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,2,2,power_law_1.01,3.494365539550781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,2,2,power_law_1.01,0.21015039443969727
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,2,2,power_law_1.01,0.2711731147766113
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,2,2,power_law_1.01,0.34959102630615235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,2,2,power_law_1.01,0.4660134506225586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,2,2,power_law_1.01,0.46642944335937503
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,2,2,power_law_1.01,0.4751628875732422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,2,2,power_law_1.01,0.4767155075073243
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,2,2,power_law_1.01,4.839455871582031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,2,2,power_law_1.01,0.50674560546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,2,2,power_law_1.01,0.5622169494628906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,2,2,power_law_1.01,0.567699203491211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,2,2,power_law_1.01,0.5853631973266602
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,2,2,power_law_1.01,0.6046771240234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,2,2,power_law_1.01,0.6019443130493164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,2,2,power_law_1.01,0.6193446350097657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,2,2,power_law_1.01,0.6659839630126954
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,2,2,power_law_1.01,6.206806030273437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,2,2,power_law_1.01,0.7310208129882813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,2,2,power_law_1.01,0.8078246307373046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,2,2,power_law_1.01,0.8987276458740234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,2,2,power_law_1.01,1.0878617858886719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,2,2,power_law_1.01,1.329559020996094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,2,2,power_law_1.2,0.09682304382324218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,2,2,power_law_1.01,1.825955810546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,2,2,power_law_1.2,0.15692543983459473
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,2,2,power_law_1.2,0.23812223434448243
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,2,2,power_law_1.01,8.809451293945312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,2,2,power_law_1.01,2.3558937072753907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,2,2,power_law_1.2,0.3746470260620117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,2,2,power_law_1.2,0.4969971084594727
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,2,2,power_law_1.2,0.6075699234008789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,2,2,power_law_1.01,3.1441421508789062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,2,2,power_law_1.2,0.9002957153320313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,2,2,power_law_1.2,0.9280831909179688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,2,2,power_law_1.2,0.9681037139892578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,2,2,power_law_1.2,0.9468224334716797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,2,2,power_law_1.2,0.9764454650878907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,2,2,power_law_1.01,11.31917724609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,2,2,power_law_1.2,1.472593994140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,2,2,power_law_1.2,1.4803071594238282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,2,2,power_law_1.2,1.569391326904297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,2,2,power_law_1.2,1.6515968322753907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,2,2,power_law_1.2,1.6365568542480466
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,2,2,power_law_1.2,1.730768585205078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,2,2,power_law_1.01,4.027415161132812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,2,2,power_law_1.2,1.915723571777344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,2,2,power_law_1.2,2.0537266540527344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,2,2,power_law_1.2,2.327151336669922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,2,2,power_law_1.2,0.06552832126617432
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,2,2,power_law_1.2,0.10499711990356446
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,2,2,power_law_1.2,3.225428466796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,2,2,power_law_1.2,0.13958016395568848
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,2,2,power_law_1.2,2.58911865234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,2,2,power_law_1.2,0.20719871520996094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,2,2,power_law_1.2,0.2595392036437988
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,2,2,power_law_1.2,0.3213529586791992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,2,2,power_law_1.2,0.45613567352294926
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,2,2,power_law_1.2,3.6925823974609373
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,2,2,power_law_1.2,0.4819379043579102
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,2,2,power_law_1.2,5.141278686523437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,2,2,power_law_1.2,0.47581825256347654
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,2,2,power_law_1.2,0.5015987014770508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,2,2,power_law_1.2,0.5578879928588867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,2,2,power_law_1.2,0.5574016189575195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,2,2,power_law_1.2,0.4714559936523438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,2,2,power_law_1.2,0.5874905776977539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,2,2,power_law_1.2,0.6016307067871094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,2,2,power_law_1.2,6.52567138671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,2,2,power_law_1.2,0.625475845336914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,2,2,power_law_1.2,0.646103057861328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,2,2,power_law_1.2,0.6901261138916015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,2,2,power_law_1.2,0.7464128112792968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,2,2,power_law_1.2,0.8796966552734375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,2,2,power_law_1.2,0.960711669921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,2,2,power_law_1.2,1.1561357116699218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,4,1,balanced,0.07761023998260498
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,2,2,power_law_1.2,1.4787481689453126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,4,1,balanced,0.10393919944763183
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,4,1,balanced,0.16830591201782225
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,4,1,balanced,0.3554668807983398
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,2,2,power_law_1.2,2.416162567138672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,2,2,power_law_1.2,9.016089477539063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,4,1,balanced,0.6015193557739258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,2,2,power_law_1.2,1.8122317504882812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,4,1,balanced,1.1359705352783203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,2,2,power_law_1.2,3.2856344604492187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,4,1,balanced,1.1444204711914063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,4,1,balanced,1.1584595489501954
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,4,1,balanced,1.2309062194824218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,4,1,balanced,1.1564288330078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,2,2,power_law_1.2,4.514561157226562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,4,1,balanced,1.1648985290527345
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,2,2,power_law_1.2,11.8232421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,4,1,balanced,1.183004150390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,4,1,balanced,1.1929529571533204
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,4,1,balanced,1.2142483520507814
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,4,1,balanced,1.775048370361328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,4,1,balanced,1.7912237548828123
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,4,1,balanced,1.8177906799316408
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,4,1,balanced,1.8912748718261718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,4,1,balanced,1.980823669433594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,4,1,balanced,2.146661834716797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,4,1,balanced,2.3004556274414063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,4,1,balanced,0.058558721542358395
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,4,1,balanced,0.076211838722229
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,4,1,balanced,0.11593600273132325
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,4,1,balanced,0.18590656280517578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,4,1,balanced,0.33839935302734375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,4,1,balanced,2.82185546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,4,1,balanced,0.6345689773559571
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,4,1,balanced,0.6391027069091797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,4,1,balanced,0.6434124755859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,4,1,balanced,2.94012939453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,4,1,balanced,0.6459839630126953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,4,1,balanced,0.6500057220458985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,4,1,balanced,4.167452697753906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,4,1,balanced,0.6568294525146484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,4,1,balanced,0.6632659149169922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,4,1,balanced,0.6673056030273438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,4,1,balanced,0.6770726776123046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,4,1,balanced,0.7615039825439454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,4,1,balanced,0.7647513580322266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,4,1,balanced,5.4586962890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,4,1,balanced,0.7894643402099609
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,4,1,balanced,0.8025228881835937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,4,1,balanced,0.756943359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,4,1,balanced,0.8636319732666016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,4,1,balanced,0.9194534301757813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,4,1,balanced,1.0225017547607422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,4,1,balanced,1.1320223999023438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,4,1,power_law_1.01,0.07643904209136962
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,4,1,power_law_1.01,0.10403008460998535
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,4,1,power_law_1.01,0.16408960342407225
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,4,1,balanced,1.7741810607910158
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,4,1,power_law_1.01,0.3555596923828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,4,1,balanced,1.9678009033203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,4,1,balanced,8.146751708984375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,4,1,power_law_1.01,0.4676780700683594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,4,1,power_law_1.01,0.6240428924560547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,4,1,balanced,2.8757684326171877
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,4,1,power_law_1.01,0.8489561462402344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,4,1,power_law_1.01,0.8858700561523438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,4,1,power_law_1.01,0.90544189453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,4,1,power_law_1.01,0.9179686737060546
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,4,1,power_law_1.01,0.9905375671386718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,4,1,balanced,3.701296691894531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,4,1,balanced,9.939794311523437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,4,1,power_law_1.01,1.0287916564941406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,4,1,power_law_1.01,1.048751983642578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,4,1,power_law_1.01,1.1058579254150391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,4,1,power_law_1.01,1.6080972290039064
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,4,1,power_law_1.01,1.621697235107422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,4,1,power_law_1.01,1.6766957092285157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,4,1,power_law_1.01,1.8065228271484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,4,1,power_law_1.01,1.9319520568847657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,4,1,power_law_1.01,2.156632995605469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,4,1,power_law_1.01,2.3800985717773435
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,4,1,power_law_1.01,0.05880000114440918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,4,1,power_law_1.01,0.07617472171783447
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,4,1,power_law_1.01,2.7916775512695313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,4,1,power_law_1.01,0.11431103706359864
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,4,1,power_law_1.01,0.18532032012939453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,4,1,power_law_1.01,0.2709139251708984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,4,1,power_law_1.01,0.34510337829589843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,4,1,power_law_1.01,3.30552001953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,4,1,power_law_1.01,0.4769171142578125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,4,1,power_law_1.01,0.4999699020385743
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,4,1,power_law_1.01,0.5114668655395508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,4,1,power_law_1.01,0.5257676696777344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,4,1,power_law_1.01,0.5484915161132813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,4,1,power_law_1.01,4.436563110351562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,4,1,power_law_1.01,0.5656684875488281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,4,1,power_law_1.01,0.5719974517822266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,4,1,power_law_1.01,0.5821625518798828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,4,1,power_law_1.01,0.6651904296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,4,1,power_law_1.01,0.6716333007812499
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,4,1,power_law_1.01,0.6981075286865235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,4,1,power_law_1.01,0.733175048828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,4,1,power_law_1.01,0.7944985961914062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,4,1,power_law_1.01,5.721793212890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,4,1,power_law_1.01,0.8906700897216797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,4,1,power_law_1.01,0.9680748748779298
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,4,1,power_law_1.01,1.1816102600097655
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,4,1,power_law_1.2,0.07734655857086181
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,4,1,power_law_1.01,1.4057369995117188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,4,1,power_law_1.2,0.10379455566406251
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,4,1,power_law_1.2,0.16502592086791992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,4,1,power_law_1.01,1.8393228149414065
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,4,1,power_law_1.2,0.3454604721069336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,4,1,power_law_1.2,0.42636543273925775
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,4,1,power_law_1.01,2.421002197265625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,4,1,power_law_1.01,8.066350097656251
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,4,1,power_law_1.2,0.5919014358520508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,4,1,power_law_1.2,0.839205093383789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,4,1,power_law_1.2,0.8697318267822265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,4,1,power_law_1.01,3.1559552001953124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,4,1,power_law_1.2,0.9304467010498048
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,4,1,power_law_1.2,0.9844166564941407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,4,1,power_law_1.2,1.0170642852783203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,4,1,power_law_1.01,3.9810977172851567
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,4,1,power_law_1.2,0.8908953857421874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,4,1,power_law_1.2,1.0638841247558593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,4,1,power_law_1.2,1.1363372802734375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,4,1,power_law_1.01,10.426322021484376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,4,1,power_law_1.2,1.6181336975097658
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,4,1,power_law_1.2,1.6493624877929687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,4,1,power_law_1.2,1.7074771118164063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,4,1,power_law_1.2,1.8828985595703123
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,4,1,power_law_1.2,1.9726451110839842
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,4,1,power_law_1.2,2.2508563232421874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,4,1,power_law_1.2,0.0588812780380249
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,4,1,power_law_1.2,2.4252032470703124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,4,1,power_law_1.2,0.07515967845916748
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,4,1,power_law_1.2,0.11207807540893555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,4,1,power_law_1.2,2.9065344238281248
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,4,1,power_law_1.2,0.1742803192138672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,4,1,power_law_1.2,0.24510847091674806
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,4,1,power_law_1.2,0.32727294921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,4,1,power_law_1.2,0.46829376220703123
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,4,1,power_law_1.2,3.43166259765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,4,1,power_law_1.2,0.4895232009887695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,4,1,power_law_1.2,0.5019379043579102
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,4,1,power_law_1.2,0.5248806381225586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,4,1,power_law_1.2,0.5454028701782226
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,4,1,power_law_1.2,0.5585849761962891
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,4,1,power_law_1.2,4.527787475585937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,4,1,power_law_1.2,0.573240966796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,4,1,power_law_1.2,0.5915385437011719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,4,1,power_law_1.2,0.6727078247070313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,4,1,power_law_1.2,0.6969580841064453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,4,1,power_law_1.2,0.712015380859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,4,1,power_law_1.2,0.7729452514648438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,4,1,power_law_1.2,5.698915405273437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,4,1,power_law_1.2,0.8549440002441406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,4,1,power_law_1.2,0.9267078399658203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,4,1,power_law_1.2,1.0162124633789062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,4,1,power_law_1.2,1.220791015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,1,4,balanced,0.021670401096343994
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,1,4,balanced,0.028157439231872562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,4,1,power_law_1.2,1.4671405029296873
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,1,4,balanced,0.03942528009414673
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,1,4,balanced,0.058617601394653315
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,1,4,balanced,0.09420415878295899
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,1,4,balanced,0.09675647735595702
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,1,4,balanced,0.12064767837524415
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,1,4,balanced,0.12089088439941406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,4,1,power_law_1.2,2.086534423828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,1,4,balanced,0.1262387180328369
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,4,1,power_law_1.2,8.005315551757812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,1,4,balanced,0.12353919982910155
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,1,4,balanced,0.1281817626953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,1,4,balanced,0.1311308765411377
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,1,4,balanced,0.1403545570373535
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,4,1,power_law_1.2,2.333738250732422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,1,4,balanced,0.13690752029418946
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,1,4,balanced,0.14854399681091307
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,1,4,balanced,0.17060991287231447
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,1,4,balanced,0.18584064483642576
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,1,4,balanced,0.1240384006500244
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,1,4,balanced,0.23948671340942385
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,1,4,balanced,0.26386175155639646
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,1,4,balanced,0.36240894317626954
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,1,4,balanced,0.45171966552734377
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,1,4,balanced,0.15028223991394044
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,1,4,balanced,0.6507981109619141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,4,1,power_law_1.2,3.182955627441406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,1,4,balanced,0.023312640190124512
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,1,4,balanced,0.024099841117858886
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,1,4,balanced,0.028247039318084716
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,1,4,balanced,0.8359744262695312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,1,4,balanced,0.04134143829345703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,1,4,balanced,0.06591360092163086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,1,4,balanced,0.06807807922363282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,1,4,balanced,0.07805439949035645
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,1,4,balanced,0.07967616081237792
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,1,4,balanced,0.08150015830993652
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,1,4,balanced,0.0844223976135254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,1,4,balanced,0.07422207832336426
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,1,4,balanced,0.07504000186920165
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,1,4,balanced,1.2296435546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,1,4,balanced,0.0804582405090332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,1,4,balanced,0.08228351593017577
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,1,4,balanced,0.0846720027923584
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,1,4,balanced,0.09063808441162109
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,1,4,balanced,0.10636799812316895
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,1,4,balanced,0.11553279876708984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,1,4,balanced,0.14079872131347657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,1,4,balanced,0.1653094482421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,1,4,balanced,1.6147686767578127
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,4,1,power_law_1.2,4.085524597167969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,1,4,balanced,0.22640512466430662
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,4,1,power_law_1.2,10.37739013671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,1,4,balanced,0.27342464447021486
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,1,4,power_law_1.01,0.037876479625701905
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,1,4,balanced,0.39402496337890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,1,4,power_law_1.01,0.04496511936187744
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,1,4,power_law_1.01,0.050159358978271486
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,1,4,balanced,0.5032601547241211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,1,4,power_law_1.01,0.05399551868438721
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,1,4,power_law_1.01,0.0626470422744751
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,1,4,power_law_1.01,0.0827724838256836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,1,4,power_law_1.01,0.09786879539489746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,1,4,power_law_1.01,0.104203519821167
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,1,4,power_law_1.01,0.10852607727050781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,1,4,power_law_1.01,0.11368320465087892
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,1,4,power_law_1.01,0.116878080368042
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,1,4,power_law_1.01,0.11709183692932128
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,1,4,power_law_1.01,0.12066944122314453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,1,4,power_law_1.01,0.13483903884887696
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,1,4,power_law_1.01,0.14344063758850098
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,1,4,power_law_1.01,0.14715392112731934
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,1,4,power_law_1.01,0.16139007568359376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,1,4,power_law_1.01,0.19439104080200195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,1,4,power_law_1.01,0.21285375595092773
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,1,4,power_law_1.01,0.2587481689453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,1,4,balanced,0.07593728065490722
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,1,4,power_law_1.01,0.3111782455444336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,1,4,balanced,0.732336654663086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,1,4,power_law_1.01,0.4282022476196289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,1,4,power_law_1.01,0.5482060623168945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,1,4,power_law_1.01,0.023697919845581054
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,1,4,power_law_1.01,0.03292160034179688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,1,4,power_law_1.01,0.03863039970397949
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,1,4,power_law_1.01,0.8258316802978516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,1,4,power_law_1.01,0.04387455940246582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,1,4,power_law_1.01,0.04924032211303711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,1,4,power_law_1.01,0.06367231845855713
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,1,4,power_law_1.01,1.0389606475830078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,1,4,power_law_1.01,0.07252863883972169
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,1,4,power_law_1.01,0.07525631904602051
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,1,4,power_law_1.01,0.08054400444030761
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,1,4,power_law_1.01,0.08194304466247558
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,1,4,power_law_1.01,0.07339136123657226
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,1,4,balanced,0.9676992034912109
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,1,4,power_law_1.01,1.5886886596679688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,1,4,power_law_1.01,0.07725952148437501
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,1,4,power_law_1.01,0.08104448318481446
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,1,4,power_law_1.01,0.08455039978027343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,1,4,power_law_1.01,0.09119999885559082
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,1,4,power_law_1.01,0.12051712036132814
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,1,4,power_law_1.01,0.15193471908569336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,1,4,power_law_1.01,0.1841996765136719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,1,4,power_law_1.01,0.24423423767089844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,1,4,power_law_1.01,0.31104127883911137
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,1,4,power_law_1.01,0.07808256149291992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,1,4,power_law_1.01,0.43820545196533206
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,1,4,power_law_1.2,0.03800575971603394
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,1,4,power_law_1.01,0.07461631774902344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,1,4,power_law_1.01,0.5697932815551757
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,1,4,power_law_1.2,0.03864959955215454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,1,4,power_law_1.2,0.04801919937133789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,1,4,power_law_1.2,0.055866880416870116
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,1,4,power_law_1.2,0.05899392127990723
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,1,4,power_law_1.01,0.10944895744323731
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,1,4,power_law_1.2,0.08204159736633301
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,1,4,power_law_1.01,0.8100351715087891
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,1,4,power_law_1.01,1.9620275878906248
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,1,4,power_law_1.2,0.10169471740722655
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,1,4,power_law_1.2,0.1087667179107666
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,1,4,power_law_1.2,0.11411328315734863
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,1,4,power_law_1.2,0.11754752159118653
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,1,4,power_law_1.2,0.1213043212890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,1,4,power_law_1.2,0.12573951721191406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,1,4,power_law_1.2,0.1362278366088867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,1,4,power_law_1.2,0.14394495964050294
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,1,4,power_law_1.2,0.15624064445495606
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,1,4,power_law_1.2,0.16830848693847655
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,1,4,power_law_1.2,0.19540735244750979
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,1,4,power_law_1.2,0.2196710395812988
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,1,4,power_law_1.2,0.2698342323303223
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,1,4,power_law_1.2,0.3358950424194336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,1,4,power_law_1.2,0.4748953628540039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,1,4,power_law_1.2,0.12415488243103028
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,1,4,power_law_1.2,0.02377216100692749
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,1,4,power_law_1.01,1.072458267211914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,1,4,power_law_1.2,0.031427841186523434
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,1,4,power_law_1.2,0.03586047887802124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,1,4,power_law_1.2,1.2366092681884766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,1,4,power_law_1.2,0.046132478713989254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,1,4,power_law_1.2,0.06420095920562743
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,1,4,power_law_1.2,0.07376383781433106
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,1,4,power_law_1.2,0.07534336090087891
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,1,4,power_law_1.2,0.07820159912109374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,1,4,power_law_1.2,0.08252415657043458
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,1,4,power_law_1.2,0.5930188751220703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,1,4,power_law_1.2,0.07476096153259278
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,1,4,power_law_1.2,0.07611648082733155
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,1,4,power_law_1.2,0.07788928031921387
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,1,4,power_law_1.2,0.08293375968933106
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,1,4,power_law_1.2,0.08649727821350098
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,1,4,power_law_1.2,0.0914675235748291
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,1,4,power_law_1.2,0.04364031791687012
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,1,4,power_law_1.2,0.11089280128479004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,1,4,power_law_1.2,0.12548992156982422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,1,4,power_law_1.2,0.8756313323974609
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,1,4,power_law_1.2,0.15370368003845214
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,1,4,power_law_1.2,0.18994176864624024
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,1,4,power_law_1.2,0.08111871719360352
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,1,4,power_law_1.2,0.3285094451904297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,1,4,power_law_1.2,0.2513983917236328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,2,2,balanced,0.018369280099868775
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,2,2,balanced,0.026123518943786624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,2,2,balanced,0.03780479907989502
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,2,2,balanced,0.05528319835662841
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,2,2,balanced,0.08868736267089844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,2,2,balanced,0.08984959602355957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,2,2,balanced,0.09054592132568359
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,2,2,balanced,0.09145855903625488
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,1,4,power_law_1.2,0.47019519805908205
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,2,2,balanced,0.11073023796081542
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,2,2,balanced,0.11063296318054198
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,2,2,balanced,0.11230336189270021
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,2,2,balanced,0.11432959556579589
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,2,2,balanced,0.11532544136047362
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,2,2,balanced,0.11789055824279786
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,2,2,balanced,0.12208000183105469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,2,2,balanced,0.12287487983703613
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,2,2,balanced,0.13783552169799804
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,2,2,balanced,0.156430082321167
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,2,2,balanced,0.1729395294189453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,2,2,balanced,0.22085632324218749
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,2,2,balanced,0.24316287994384767
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,2,2,balanced,0.3422963333129883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,2,2,balanced,0.43016193389892576
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,2,2,balanced,0.6196940612792969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,1,4,power_law_1.2,1.808001251220703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,2,2,balanced,0.7966809844970703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,2,2,balanced,0.021660161018371583
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,2,2,balanced,0.021822719573974608
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,2,2,balanced,0.02622463941574097
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,2,2,balanced,0.04290304183959961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,2,2,balanced,0.06596608161926269
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,2,2,balanced,0.06785664081573486
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,2,2,balanced,0.06818687915802002
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,1,4,power_law_1.2,0.5609126281738281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,2,2,balanced,0.06853504180908203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,2,2,balanced,0.07851903915405274
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,2,2,balanced,0.07884672164916992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,2,2,balanced,0.07976831912994384
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,2,2,balanced,0.08095104217529296
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,2,2,balanced,1.5395033264160156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,2,2,balanced,0.08116736412048339
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,2,2,balanced,0.08330752372741698
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,2,2,balanced,0.08548352241516113
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,2,2,balanced,0.08690815925598144
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,2,2,balanced,0.10252799987792968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,2,2,balanced,0.11379455566406249
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,2,2,balanced,0.13209343910217286
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,2,2,balanced,0.15297023773193358
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,2,2,balanced,1.1748505401611329
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,2,2,balanced,0.2181644821166992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,2,2,balanced,0.254833927154541
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,2,2,balanced,0.37430015563964847
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,2,2,balanced,0.47754367828369143
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,2,2,power_law_1.01,0.020614399909973144
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,2,2,power_law_1.01,0.03285120010375976
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,2,2,balanced,0.6978739166259766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,2,2,power_law_1.01,0.04116991996765137
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,2,2,power_law_1.01,0.04695551872253418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,2,2,power_law_1.01,0.0575705623626709
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,2,2,power_law_1.01,0.07456255912780761
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,2,2,balanced,0.9110860443115234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,2,2,power_law_1.01,0.08144895553588867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,2,2,power_law_1.01,0.08319104194641114
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,2,2,power_law_1.01,0.0980684757232666
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,2,2,power_law_1.01,0.10071552276611329
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,2,2,power_law_1.01,0.10495615959167481
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,2,2,power_law_1.01,0.1081497573852539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,2,2,power_law_1.01,0.11046272277832032
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,2,2,power_law_1.01,0.11634559631347656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,2,2,power_law_1.01,0.1252403163909912
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,2,2,power_law_1.01,0.13328000068664553
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,2,2,power_law_1.01,0.14668031692504885
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,2,2,balanced,0.09142656326293945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,2,2,power_law_1.01,0.18236799240112306
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,2,2,power_law_1.01,0.19637887954711913
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,2,2,power_law_1.01,0.22997503280639647
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,2,2,power_law_1.01,0.2827097511291504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,2,2,power_law_1.01,0.3752729415893555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,1,4,power_law_1.2,0.8451827239990234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,2,2,power_law_1.01,0.47829761505126955
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,2,2,power_law_1.01,0.02164223909378052
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,2,2,power_law_1.01,0.6872051239013672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,2,2,power_law_1.01,0.02602880001068115
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,2,2,power_law_1.01,0.03381376028060913
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,2,2,power_law_1.01,0.038613760471343996
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,2,2,power_law_1.01,0.8933542633056641
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,2,2,power_law_1.01,0.04540544033050537
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,2,2,power_law_1.01,0.06154496192932128
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,2,2,power_law_1.01,0.06387584209442139
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,2,2,power_law_1.01,0.06439040184020997
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,2,2,power_law_1.01,0.07539455890655518
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,2,2,power_law_1.01,0.07677055835723877
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,2,2,power_law_1.01,0.07793407917022706
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,2,2,power_law_1.01,0.07903232097625731
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,2,2,power_law_1.01,1.3186572265624998
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,2,2,power_law_1.01,0.08104191780090332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,2,2,power_law_1.01,0.08388992309570312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,2,2,power_law_1.01,0.08554368019104004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,2,2,power_law_1.01,0.08739199638366699
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,2,2,power_law_1.01,0.09257599830627442
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,2,2,power_law_1.01,0.10692607879638671
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,2,2,power_law_1.01,0.11883008003234863
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,2,2,power_law_1.01,0.15074432373046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,2,2,power_law_1.01,1.7379954528808592
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,2,2,power_law_1.01,0.17653888702392578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,2,2,power_law_1.01,0.23668479919433594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,2,2,power_law_1.01,0.2923750305175781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,2,2,power_law_1.01,0.4077619171142578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,2,2,power_law_1.2,0.02067456007003784
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,2,2,power_law_1.01,0.5301824188232421
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,2,2,power_law_1.2,0.031374080181121825
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,2,2,power_law_1.2,0.03933439970016479
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,2,2,power_law_1.01,0.756421127319336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,2,2,power_law_1.2,0.0422105598449707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,2,2,power_law_1.2,0.049538559913635254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,2,2,power_law_1.2,0.07552383899688721
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,2,2,power_law_1.2,0.0848793601989746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,2,2,power_law_1.2,0.1029695987701416
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,2,2,power_law_1.01,0.9787506866455079
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,2,2,power_law_1.2,0.10635007858276366
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,2,2,power_law_1.2,0.11095295906066896
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,2,2,power_law_1.2,0.11066752433776857
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,2,2,power_law_1.2,0.11525119781494139
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,2,2,power_law_1.2,0.12180352210998535
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,2,2,power_law_1.2,0.12640000343322755
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,2,2,power_law_1.2,0.13431296348571778
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,2,2,power_law_1.2,0.14622336387634277
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,2,2,power_law_1.2,0.18271999359130858
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,2,2,power_law_1.2,0.1967987251281738
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,2,2,power_law_1.2,0.23379327774047853
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,2,2,power_law_1.2,0.29198335647583007
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,2,2,power_law_1.2,0.07998079776763915
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,2,2,power_law_1.2,0.4004352188110351
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,2,2,power_law_1.2,0.021351680755615235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,2,2,power_law_1.2,0.9675007629394532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,2,2,power_law_1.2,0.0258790397644043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,2,2,power_law_1.2,0.03175807952880859
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,2,2,power_law_1.2,0.035777280330657954
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,2,2,power_law_1.2,0.04068096160888672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,2,2,power_law_1.2,0.06221695899963379
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,2,2,power_law_1.2,0.5005721664428711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,2,2,power_law_1.2,0.06409215927124023
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,2,2,power_law_1.2,0.06531839847564698
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,2,2,power_law_1.2,0.07607295989990234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,2,2,power_law_1.2,0.0772544002532959
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,2,2,power_law_1.2,0.07995776176452637
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,2,2,power_law_1.2,0.08141695976257324
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,2,2,power_law_1.2,0.08372991561889649
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,2,2,power_law_1.2,0.712273941040039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,2,2,power_law_1.2,0.08666367530822754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,2,2,power_law_1.2,0.08865407943725587
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,2,2,power_law_1.2,0.09351679801940918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,2,2,power_law_1.2,0.10800000190734864
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,2,2,power_law_1.2,0.1256985569000244
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,2,2,power_law_1.2,0.15067392349243164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,2,2,power_law_1.2,0.07870079994201659
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,2,2,power_law_1.2,0.17932800292968748
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,2,2,power_law_1.2,0.29559167861938473
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,2,2,power_law_1.2,0.23692031860351564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,1,4,power_law_1.2,1.0884339141845705
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,2,2,power_law_1.2,1.3844108581542969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,2,2,power_law_1.2,0.41693183898925784
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,2,2,power_law_1.2,0.5286463928222657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,4,1,balanced,0.02515775918960571
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,4,1,balanced,0.04359360218048096
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,4,1,balanced,0.03396608114242554
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,2,2,power_law_1.2,0.7691391754150391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,4,1,balanced,0.060730881690979
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,4,1,balanced,0.09583488464355469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,4,1,balanced,0.09899776458740235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,4,1,balanced,0.09958784103393555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,4,1,balanced,0.10036031723022461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,4,1,balanced,0.10506431579589844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,4,1,balanced,0.11273088455200195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,4,1,balanced,0.11512127876281739
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,4,1,balanced,0.1168633556365967
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,4,1,balanced,0.1028012752532959
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,1,4,power_law_1.2,2.398303985595703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,4,1,balanced,0.10695103645324706
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,4,1,balanced,0.1597740840911865
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,4,1,balanced,0.1750668716430664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,4,1,balanced,0.10845184326171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,4,1,balanced,0.10970879554748536
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,2,2,power_law_1.2,0.9907328033447266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,4,1,balanced,0.12283072471618653
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,4,1,balanced,0.1304044818878174
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,4,1,balanced,0.2728940773010254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,4,1,balanced,0.30506048202514646
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,4,1,balanced,0.20627328872680667
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,4,1,balanced,0.43878974914550783
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,4,1,balanced,0.5632902526855469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,4,1,balanced,0.8245823669433594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,4,1,balanced,0.026078720092773438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,4,1,balanced,0.027541120052337647
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,4,1,balanced,1.0725663757324218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,4,1,balanced,0.0348691201210022
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,4,1,balanced,0.05056128025054931
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,4,1,balanced,0.06942207813262939
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,4,1,balanced,0.07066559791564941
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,4,1,balanced,0.07711872100830078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,4,1,balanced,0.07842239856719971
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,4,1,balanced,0.10068544387817382
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,4,1,balanced,0.10201087951660157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,2,2,power_law_1.2,1.816302032470703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,4,1,balanced,0.07138239860534668
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,4,1,balanced,0.07290495872497559
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,4,1,balanced,0.07321663856506347
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,4,1,balanced,0.07402944087982177
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,4,1,balanced,0.07540224075317384
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,4,1,balanced,0.0806553554534912
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,4,1,balanced,0.09152959823608399
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,4,1,balanced,0.09856575965881348
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,4,1,balanced,0.16528192520141602
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,4,1,balanced,0.22392959594726564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,4,1,balanced,0.13879103660583497
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,4,1,balanced,0.10721407890319826
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,4,1,power_law_1.01,0.04350399971008301
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,4,1,balanced,0.2551968002319336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,4,1,power_law_1.01,0.025011200904846192
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,4,1,balanced,0.3673702239990234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,4,1,balanced,0.4747328186035157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,4,1,power_law_1.01,0.05353856086730957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,4,1,balanced,0.6944966125488281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,4,1,power_law_1.01,0.03411328077316284
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,4,1,power_law_1.01,0.11381888389587402
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,4,1,power_law_1.01,0.11623552322387695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,4,1,power_law_1.01,0.11863231658935547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,4,1,power_law_1.01,0.06319551944732667
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,4,1,balanced,0.9145299530029296
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,4,1,power_law_1.01,0.09220671653747559
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,4,1,power_law_1.01,0.09574272155761718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,4,1,power_law_1.01,0.08750016212463378
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,4,1,power_law_1.01,0.1402732753753662
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,4,1,power_law_1.01,0.18706432342529297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,4,1,power_law_1.01,0.21032768249511719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,4,1,power_law_1.01,0.09731648445129396
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,4,1,power_law_1.01,0.09743871688842773
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,4,1,power_law_1.01,0.12782015800476074
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,4,1,power_law_1.01,0.13027520179748536
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,4,1,power_law_1.01,0.13329471588134764
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,4,1,power_law_1.01,0.6207308959960938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,4,1,power_law_1.01,0.8880684661865235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,4,1,power_law_1.01,1.1470758056640624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,4,1,power_law_1.01,0.14788864135742188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,4,1,power_law_1.01,0.22460607528686521
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,4,1,power_law_1.01,0.3008153533935547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,4,1,power_law_1.01,0.3576268768310547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,4,1,power_law_1.01,0.0643558406829834
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,4,1,power_law_1.01,0.49070209503173834
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,4,1,power_law_1.01,0.06920063972473145
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,4,1,power_law_1.01,0.07097472190856934
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,4,1,power_law_1.01,0.07133312225341797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,4,1,power_law_1.01,0.0767257595062256
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,4,1,power_law_1.01,0.07888192176818848
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,4,1,power_law_1.01,0.08435775756835937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,4,1,power_law_1.01,0.0985267162322998
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,4,1,power_law_1.01,0.10321920394897462
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,4,1,power_law_1.01,0.025855360031127927
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,4,1,power_law_1.01,0.09226559638977051
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,4,1,power_law_1.01,0.028492159843444824
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,4,1,power_law_1.01,0.03805504083633423
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,4,1,power_law_1.01,0.17809791564941407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,4,1,power_law_1.01,0.05067584037780761
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,4,1,power_law_1.01,0.04468095779418945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,4,1,power_law_1.01,0.06806911945343017
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,4,1,power_law_1.01,0.5133670425415039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,4,1,power_law_1.01,0.733986587524414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,4,1,power_law_1.01,0.11341888427734374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,4,1,power_law_1.01,0.1009055995941162
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,4,1,power_law_1.2,0.02532416105270386
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,4,1,power_law_1.01,0.9549964904785156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,4,1,power_law_1.01,0.16421247482299806
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,4,1,power_law_1.01,0.12413375854492188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,4,1,power_law_1.2,0.06074048042297363
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,4,1,power_law_1.2,0.09744768142700196
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,4,1,power_law_1.01,0.2442982482910156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,4,1,power_law_1.01,0.2922604751586914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,4,1,power_law_1.01,0.40561279296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,4,1,power_law_1.2,0.11407872200012208
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,4,1,power_law_1.2,0.05144896030426025
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,4,1,power_law_1.2,0.03482239961624146
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,4,1,power_law_1.2,0.1279321575164795
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,4,1,power_law_1.2,0.04279551982879638
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,4,1,power_law_1.2,0.13033087730407716
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,4,1,power_law_1.2,0.1328998374938965
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,4,1,power_law_1.2,0.11762559890747071
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,4,1,power_law_1.2,0.09455103874206543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,4,1,power_law_1.2,0.1889094352722168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,4,1,power_law_1.2,0.09722111701965333
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,4,1,power_law_1.2,0.0876748752593994
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,4,1,power_law_1.2,0.21903295516967775
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,4,1,power_law_1.2,0.09889408111572265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,4,1,power_law_1.2,0.36797374725341797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,4,1,power_law_1.2,0.11996095657348634
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,4,1,power_law_1.2,0.13935232162475586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,4,1,power_law_1.2,0.14845439910888672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,4,1,power_law_1.2,0.49761280059814456
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,4,1,power_law_1.2,0.030381441116333008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,4,1,power_law_1.2,0.03716928005218506
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,4,1,power_law_1.2,0.2252300834655762
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,4,1,power_law_1.2,0.30105791091918943
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,4,1,power_law_1.2,0.06460351943969726
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,4,1,power_law_1.2,0.06835968017578124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,4,1,power_law_1.2,0.06993855953216552
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,4,1,power_law_1.2,0.04240128040313721
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,4,1,power_law_1.2,0.07141056060791015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,4,1,power_law_1.2,0.07260608196258544
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,4,1,power_law_1.2,0.07673215866088867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,4,1,power_law_1.2,0.07971456050872802
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,4,1,power_law_1.2,0.08584896087646485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,4,1,power_law_1.2,0.09294719696044922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,4,1,power_law_1.2,0.09889023780822755
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,4,1,power_law_1.2,0.10419136047363282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,4,1,power_law_1.2,0.10241472244262695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,4,1,power_law_1.2,0.16722431182861328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,4,1,power_law_1.2,0.025804800987243654
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,4,1,power_law_1.2,0.04857600212097168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,4,1,power_law_1.2,0.6251494216918945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,4,1,power_law_1.2,0.2979225540161133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,4,1,power_law_1.2,0.11460800170898437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,4,1,power_law_1.2,0.17715200424194336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,4,1,power_law_1.2,0.12530943870544434
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,1,4,balanced,0.05115647792816162
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,1,4,balanced,0.06184447765350341
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,1,4,balanced,0.09304960250854492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,1,4,balanced,0.17535999298095703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,4,1,power_law_1.2,0.40783039093017576
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,1,4,balanced,0.3006182479858398
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,4,1,power_law_1.2,0.2443564796447754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,1,4,balanced,0.30476800918579106
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,1,4,balanced,0.45088638305664064
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,1,4,balanced,0.456677131652832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,1,4,balanced,0.46171520233154295
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,1,4,balanced,0.47287937164306637
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,1,4,balanced,0.4675788879394531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,1,4,balanced,0.4731955337524414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,1,4,balanced,0.47752830505371097
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,1,4,balanced,0.4865702438354492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,1,4,balanced,0.5089766311645507
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,1,4,balanced,0.5175244903564453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,1,4,balanced,0.5431206512451172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,1,4,balanced,0.5732928085327148
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,1,4,balanced,0.6090892791748047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,1,4,balanced,0.717817611694336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,1,4,balanced,0.808564453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,1,4,balanced,1.136363525390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,1,4,balanced,1.4267878723144531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,1,4,balanced,2.109966125488281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,1,4,balanced,2.7004583740234374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,1,4,balanced,0.032060160636901855
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,4,1,power_law_1.2,0.5166041564941406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,1,4,balanced,0.04200831890106201
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,1,4,balanced,0.05917823791503907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,1,4,balanced,0.0990886402130127
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,1,4,balanced,0.16979583740234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,1,4,balanced,0.17291519165039063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,1,4,balanced,0.19421056747436521
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,1,4,balanced,0.19753856658935548
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,1,4,balanced,0.19970687866210937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,1,4,balanced,0.20243455886840822
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,1,4,balanced,4.0419775390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,1,4,balanced,0.19425151824951173
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,1,4,balanced,0.19551231384277343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,1,4,balanced,0.19807743072509765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,1,4,balanced,0.2010304069519043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,1,4,balanced,0.20644479751586914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,1,4,balanced,0.2114419174194336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,1,4,balanced,0.22421247482299805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,1,4,balanced,0.2537215995788574
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,1,4,balanced,0.2685862350463867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,1,4,balanced,0.31124607086181644
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,1,4,balanced,0.35806079864501955
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,1,4,balanced,0.4957683181762695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,1,4,balanced,0.5846464157104492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,1,4,balanced,0.8424230194091796
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,1,4,balanced,5.469083862304688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,1,4,power_law_1.01,0.08834943771362305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,1,4,balanced,1.073798370361328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,4,1,power_law_1.2,0.8897516632080078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,1,4,power_law_1.01,0.10408320426940917
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,1,4,power_law_1.01,0.15245439529418944
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,1,4,power_law_1.01,0.1625984001159668
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,1,4,power_law_1.01,0.19892223358154298
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,1,4,balanced,1.6028236389160155
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,1,4,power_law_1.01,0.4146176147460937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,1,4,power_law_1.01,0.4284262466430664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,1,4,power_law_1.01,0.4488576126098633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,1,4,balanced,2.0745689392089846
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,1,4,power_law_1.01,0.45613441467285154
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,1,4,power_law_1.01,0.4601011276245117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,1,4,power_law_1.01,0.4734809494018554
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,1,4,power_law_1.01,0.46720897674560546
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,1,4,power_law_1.01,0.5093324661254883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,1,4,power_law_1.01,0.5158425521850586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,1,4,power_law_1.01,0.5362022399902344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,1,4,power_law_1.01,0.5602700805664063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,1,4,power_law_1.01,0.6169561767578126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,1,4,power_law_1.01,0.7322509002685547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,1,4,power_law_1.01,0.9170496368408203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,1,4,power_law_1.01,1.0124877166748045
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,1,4,power_law_1.01,0.2818547248840332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,1,4,power_law_1.01,1.4611059570312501
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,1,4,power_law_1.01,1.958271942138672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,1,4,power_law_1.01,2.6592587280273436
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,1,4,power_law_1.01,0.05758975982666016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,1,4,power_law_1.01,0.0705024003982544
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,1,4,power_law_1.01,0.09267840385437012
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,1,4,power_law_1.01,0.10466048240661621
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,1,4,power_law_1.01,0.1179481601715088
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,1,4,power_law_1.01,0.16067455291748048
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,1,4,power_law_1.01,0.18885631561279298
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,1,4,power_law_1.01,0.18436992645263672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,1,4,power_law_1.01,0.19325439453124998
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,1,4,power_law_1.01,0.19349376678466795
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,1,4,power_law_1.01,0.19785728454589843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,1,4,power_law_1.01,3.4823193359374995
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,1,4,power_law_1.01,0.19017984390258788
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,1,4,power_law_1.01,0.18888320922851562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,1,4,power_law_1.01,0.19536640167236327
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,1,4,power_law_1.01,0.20382463455200198
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,1,4,power_law_1.01,0.21177600860595702
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,1,4,power_law_1.01,0.227457275390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,1,4,power_law_1.01,0.2683686447143555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,1,4,power_law_1.01,0.28462335586547854
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,1,4,power_law_1.01,0.350937614440918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,1,4,power_law_1.01,0.42471038818359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,1,4,power_law_1.01,0.5545702362060547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,1,4,power_law_1.01,0.6917542266845703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,1,4,power_law_1.01,5.61491455078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,1,4,power_law_1.01,0.9322291564941405
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,4,1,power_law_1.2,0.7390393829345703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,1,4,power_law_1.2,0.08650879859924317
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,1,4,power_law_1.2,0.1028275203704834
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,1,4,power_law_1.01,1.295160369873047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,1,4,power_law_1.2,0.1143603229522705
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,1,4,power_law_1.2,0.15824640274047852
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,1,4,power_law_1.2,0.22580480575561523
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,1,4,power_law_1.2,0.28480512619018555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,1,4,power_law_1.01,1.8865216064453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,1,4,power_law_1.2,0.4178841781616211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,1,4,power_law_1.2,0.4379980850219726
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,1,4,power_law_1.2,0.4513638305664062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,1,4,power_law_1.2,0.4541094589233398
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,1,4,power_law_1.2,0.4702259063720703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,1,4,power_law_1.01,7.165375976562499
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,1,4,power_law_1.2,0.48081279754638667
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,1,4,power_law_1.2,0.5188825607299805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,1,4,power_law_1.01,2.6673843383789064
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,1,4,power_law_1.2,0.5140428924560547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,1,4,power_law_1.2,0.5228006362915039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,1,4,power_law_1.2,0.5335257720947266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,1,4,power_law_1.2,0.5807462310791015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,1,4,power_law_1.2,0.6318783950805664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,1,4,power_law_1.2,0.6896051025390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,1,4,power_law_1.2,1.0073458862304687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,1,4,power_law_1.2,1.0505433654785157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,1,4,power_law_1.2,1.612093505859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,1,4,power_law_1.2,0.05805056095123291
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,1,4,power_law_1.2,0.0720358419418335
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,1,4,power_law_1.2,0.08154623985290528
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,1,4,power_law_1.2,0.09986944198608398
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,1,4,power_law_1.2,0.12011263847351075
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,1,4,power_law_1.2,2.2782579040527344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,1,4,power_law_1.2,0.15923328399658204
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,1,4,power_law_1.2,0.18213375091552736
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,1,4,power_law_1.2,0.18994943618774413
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,1,4,power_law_1.2,0.19001728057861328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,1,4,power_law_1.2,3.1492404174804687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,1,4,power_law_1.2,0.1939289665222168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,1,4,power_law_1.2,0.19719167709350588
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,1,4,power_law_1.2,0.19529600143432618
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,1,4,power_law_1.2,0.1887334442138672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,1,4,power_law_1.2,0.19625856399536132
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,1,4,power_law_1.2,0.21327743530273438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,1,4,power_law_1.2,0.21669631958007812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,1,4,power_law_1.2,0.2344076728820801
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,1,4,power_law_1.2,0.2679206466674805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,1,4,power_law_1.2,0.2944870376586914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,1,4,power_law_1.2,0.35938945770263675
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,1,4,power_law_1.2,0.43008895874023434
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,1,4,power_law_1.2,0.5889740753173828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,1,4,power_law_1.2,0.7148287963867188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,4,1,power_law_1.2,1.1526451110839844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,1,4,power_law_1.2,3.861827697753906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,1,4,power_law_1.2,1.0792639923095702
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,2,2,balanced,0.04796544075012207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,2,2,balanced,0.06125951766967773
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,2,2,balanced,0.09644031524658203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,2,2,balanced,0.1631987190246582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,2,2,balanced,0.2982156753540039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,2,2,balanced,0.2996185684204101
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,2,2,balanced,0.30449920654296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,2,2,balanced,0.3058252716064453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,2,2,balanced,0.4372364807128906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,2,2,balanced,0.44355201721191406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,2,2,balanced,0.4467212677001953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,2,2,balanced,0.45548286437988283
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,2,2,balanced,0.45880702972412113
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,2,2,balanced,0.47458049774169925
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,2,2,balanced,0.48661247253417966
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,2,2,balanced,0.4994675064086914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,2,2,balanced,0.5245094299316406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,2,2,balanced,0.5525388717651367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,2,2,balanced,0.5821068954467774
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,2,2,balanced,0.6961344146728516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,2,2,balanced,0.7890188598632812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,1,4,power_law_1.2,6.511133422851563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,2,2,balanced,1.0982067108154296
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,1,4,power_law_1.2,1.3411276245117187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,2,2,balanced,1.3939724731445313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,2,2,balanced,2.0359948730468753
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,2,2,balanced,0.029922559261322024
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,2,2,balanced,0.042007040977478025
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,2,2,balanced,2.617479553222656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,2,2,balanced,0.06128128051757813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,2,2,balanced,0.10137344360351563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,2,2,balanced,0.17257471084594728
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,2,2,balanced,0.17219583511352538
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,2,2,balanced,0.17506431579589843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,2,2,balanced,0.17632383346557617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,2,2,balanced,0.19484159469604492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,2,2,balanced,0.19554943084716797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,2,2,balanced,0.1964031982421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,2,2,balanced,0.198919677734375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,2,2,balanced,0.20060543060302732
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,2,2,balanced,0.20313215255737305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,2,2,balanced,0.20648448944091796
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,2,2,balanced,0.21024896621704103
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,2,2,balanced,0.21883520126342773
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,2,2,balanced,0.23746175765991212
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,2,2,balanced,3.8924530029296873
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,2,2,balanced,0.2616883277893066
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,2,2,balanced,0.28998271942138676
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,2,2,balanced,5.051415100097656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,2,2,balanced,0.33281665802001953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,2,2,balanced,0.48305534362792973
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,2,2,balanced,0.5530815887451171
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,2,2,balanced,0.7985497283935546
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,2,2,power_law_1.01,0.05894144058227539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,2,2,balanced,1.0369446563720703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,2,2,power_law_1.01,0.07464064121246337
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,2,2,power_law_1.01,0.1089151954650879
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,2,2,power_law_1.01,0.13387392044067384
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,2,2,power_law_1.01,0.16289024353027343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,2,2,power_law_1.01,0.2635840034484863
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,2,2,power_law_1.01,0.2719027137756348
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,2,2,balanced,2.0127821350097657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,2,2,power_law_1.01,0.2769011116027832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,2,2,power_law_1.01,0.41862400054931637
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,2,2,power_law_1.01,0.4277503967285156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,2,2,balanced,1.5183360290527343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,2,2,power_law_1.01,0.4318783950805664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,2,2,power_law_1.01,0.45251201629638677
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,2,2,power_law_1.01,0.45852031707763674
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,2,2,power_law_1.01,0.4815654373168945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,2,2,power_law_1.01,0.4927347183227539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,2,2,power_law_1.01,0.5177484893798828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,2,2,power_law_1.01,0.576200942993164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,2,2,power_law_1.01,0.6332825469970703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,2,2,power_law_1.01,0.46884609222412105
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,2,2,power_law_1.01,0.7767718505859376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,4,1,power_law_1.2,0.9537535858154296
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,2,2,power_law_1.01,0.9191129302978516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,2,2,power_law_1.01,1.262118377685547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,2,2,power_law_1.01,1.6813325500488283
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,2,2,power_law_1.01,0.04129792213439941
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,2,2,power_law_1.01,0.05370240211486817
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,2,2,power_law_1.01,2.3515942382812502
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,2,2,power_law_1.01,0.07085440158843995
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,2,2,power_law_1.01,0.09091327667236329
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,2,2,power_law_1.01,2.7924172973632815
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,2,2,power_law_1.01,0.11194368362426758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,2,2,power_law_1.01,0.15868288040161133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,2,2,power_law_1.01,0.15597311973571776
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,2,2,power_law_1.01,0.18523391723632812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,2,2,power_law_1.01,0.16516992568969727
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,2,2,power_law_1.01,0.18594175338745117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,2,2,power_law_1.01,0.1930112075805664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,2,2,power_law_1.01,0.19484159469604492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,2,2,power_law_1.01,0.20156799316406251
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,2,2,power_law_1.01,0.20824960708618162
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,2,2,power_law_1.01,0.20380800247192382
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,1,4,power_law_1.2,1.919280700683594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,2,2,power_law_1.01,4.529456787109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,2,2,power_law_1.01,0.19300735473632813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,2,2,power_law_1.01,0.22597120285034178
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,2,2,power_law_1.01,0.25575551986694334
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,2,2,power_law_1.01,0.2834048080444336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,2,2,power_law_1.01,0.3463718414306641
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,2,2,power_law_1.01,0.40306816101074217
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,2,2,power_law_1.01,0.5249087905883789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,2,2,power_law_1.01,0.6439142608642578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,2,2,power_law_1.2,0.05882880210876464
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,2,2,power_law_1.01,0.8917862701416016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,2,2,power_law_1.2,0.06941823959350586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,2,2,power_law_1.2,0.10144895553588866
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,2,2,power_law_1.01,5.874507446289063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,2,2,power_law_1.2,0.1164851188659668
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,2,2,power_law_1.2,0.15064831733703615
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,2,2,power_law_1.2,0.27383424758911135
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,2,2,power_law_1.01,1.7451405334472656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,2,2,power_law_1.2,0.27620735168457033
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,2,2,power_law_1.2,0.2881868743896484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,2,2,power_law_1.01,1.1859468841552734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,2,2,power_law_1.2,0.4330598449707031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,2,2,power_law_1.01,2.2412646484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,2,2,power_law_1.2,0.44804222106933594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,2,2,power_law_1.2,0.4481062316894532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,2,2,power_law_1.2,0.4555942535400391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,2,2,power_law_1.2,0.4704806518554688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,2,2,power_law_1.2,0.4876492691040039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,2,2,power_law_1.2,0.4840768051147461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,2,2,power_law_1.2,0.5220915222167969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,2,2,power_law_1.2,0.43639167785644534
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,2,2,power_law_1.2,0.5868249511718749
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,2,2,power_law_1.2,0.6505420684814454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,2,2,power_law_1.2,0.8123993682861329
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,2,2,power_law_1.2,0.9215974426269531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,2,2,power_law_1.2,1.2882290649414063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,2,2,power_law_1.2,0.04141568183898926
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,2,2,power_law_1.2,0.05128191947937012
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,2,2,power_law_1.2,0.0653273582458496
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,2,2,power_law_1.2,2.463233337402344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,1,4,power_law_1.2,2.891089782714844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,1,4,power_law_1.2,8.36550537109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,2,2,power_law_1.2,0.08687104225158691
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,2,2,power_law_1.2,0.09841919898986816
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,2,2,power_law_1.2,0.15949055671691895
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,2,2,power_law_1.2,0.15881216049194335
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,2,2,power_law_1.2,0.1673151969909668
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,2,2,power_law_1.2,1.7189759826660158
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,2,2,power_law_1.2,0.18717952728271486
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,2,2,power_law_1.2,0.19175039291381837
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,2,2,power_law_1.2,0.19588607788085938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,2,2,power_law_1.2,0.19495040893554688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,2,2,power_law_1.2,0.19932928085327148
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,2,2,power_law_1.2,0.1994316864013672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,2,2,power_law_1.2,0.21270143508911135
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,2,2,power_law_1.2,0.234521598815918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,2,2,power_law_1.2,0.2564633560180664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,2,2,power_law_1.2,0.2850662422180176
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,2,2,power_law_1.2,0.3465254211425781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,2,2,power_law_1.2,0.40614910125732423
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,2,2,power_law_1.2,0.2121932792663574
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,2,2,power_law_1.2,0.5340326309204102
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,2,2,power_law_1.2,0.65233154296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,2,2,power_law_1.2,0.9314739227294921
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,2,2,power_law_1.2,2.9614990234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,4,1,balanced,0.09722047805786133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,4,1,balanced,0.1642201614379883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,4,1,balanced,0.048311681747436525
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,4,1,balanced,0.06276159763336182
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,2,2,power_law_1.2,1.2056755065917968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,4,1,balanced,0.32075393676757813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,4,1,balanced,0.30727615356445315
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,4,1,balanced,0.31997631072998045
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,2,2,power_law_1.2,4.901987915039062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,4,1,balanced,0.3152518463134766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,4,1,balanced,0.31880319595336915
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,4,1,balanced,0.3231123352050781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,4,1,balanced,0.32329856872558593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,4,1,balanced,0.3385728073120117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,4,1,balanced,0.31424703598022463
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,4,1,balanced,0.3515660858154297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,4,1,balanced,0.36347713470458987
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,4,1,balanced,0.41077438354492185
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,4,1,balanced,0.4397990417480469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,4,1,balanced,0.3088006401062012
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,4,1,balanced,0.4788320159912109
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,4,1,balanced,0.7192704010009765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,4,1,balanced,0.774730224609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,4,1,balanced,1.1398969268798829
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,4,1,balanced,1.482042236328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,4,1,balanced,0.036662399768829346
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,4,1,balanced,0.05139455795288086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,4,1,balanced,0.07189824104309081
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,4,1,balanced,2.215370941162109
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,4,1,balanced,0.1157875156402588
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,4,1,balanced,0.19444480895996094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,4,1,balanced,0.1977248001098633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,4,1,balanced,0.1979110336303711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,4,1,balanced,0.19914495468139648
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,4,1,balanced,0.20089855194091796
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,4,1,balanced,0.33972991943359376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,4,1,balanced,0.34317249298095703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,4,1,balanced,0.20145856857299807
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,4,1,balanced,0.20327104568481444
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,4,1,balanced,2.8924166870117185
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,4,1,balanced,0.23173696517944337
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,4,1,balanced,0.23293119430541992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,4,1,balanced,0.23740159988403323
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,4,1,balanced,0.2402272033691406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,4,1,balanced,0.24569984436035158
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,4,1,balanced,0.24742591857910154
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,4,1,balanced,0.2603238487243652
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,4,1,balanced,0.27478656768798826
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,4,1,balanced,0.30547712326049803
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,4,1,balanced,0.33730369567871094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,4,1,balanced,0.5044915390014648
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,4,1,balanced,0.5658803176879883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,4,1,balanced,0.882305908203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,4,1,balanced,1.0856966400146484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,4,1,power_law_1.01,0.0507539176940918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,4,1,power_law_1.01,0.06234879970550537
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,4,1,balanced,1.589842529296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,4,1,balanced,2.091200714111328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,4,1,power_law_1.01,0.2581113624572754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,4,1,power_law_1.01,0.2790592002868652
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,4,1,power_law_1.01,0.2908448028564453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,4,1,power_law_1.01,0.30192447662353517
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,4,1,power_law_1.01,0.29641664505004883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,4,1,power_law_1.01,0.3204083251953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,2,2,power_law_1.2,1.6984538269042968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,4,1,power_law_1.01,0.3309292984008789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,4,1,power_law_1.01,0.34095550537109376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,4,1,power_law_1.01,0.09605824470520019
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,4,1,power_law_1.01,0.3451129531860352
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,4,1,power_law_1.01,0.3377305603027344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,4,1,power_law_1.01,0.3679948806762695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,4,1,power_law_1.01,0.3951468658447266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,4,1,power_law_1.01,0.44583297729492183
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,4,1,power_law_1.01,0.13416128158569335
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,4,1,power_law_1.01,0.5315667343139648
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,4,1,power_law_1.01,0.38605697631835934
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,4,1,power_law_1.01,0.627325439453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,4,1,power_law_1.01,0.7765586853027344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,4,1,power_law_1.01,0.9456800079345703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,4,1,power_law_1.01,0.17254655838012695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,4,1,power_law_1.01,0.03427072048187256
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,4,1,power_law_1.01,0.05067520141601563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,4,1,power_law_1.01,1.3098585510253906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,4,1,power_law_1.01,0.07032832145690918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,4,1,power_law_1.01,0.1183232021331787
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,4,1,power_law_1.01,0.16926080703735352
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,4,1,power_law_1.01,1.6641920471191405
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,4,1,power_law_1.01,0.1787295913696289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,4,1,power_law_1.01,0.18328128814697267
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,4,1,power_law_1.01,0.18959104537963867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,4,1,power_law_1.01,0.18846656799316405
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,4,1,power_law_1.01,0.10198016166687011
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,4,1,power_law_1.01,0.19562751770019532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,4,1,power_law_1.01,2.3883021545410155
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,4,1,power_law_1.01,0.22754560470581056
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,4,1,power_law_1.01,0.23394432067871093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,4,1,power_law_1.01,0.24135808944702147
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,4,1,power_law_1.01,0.24743167877197267
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,4,1,power_law_1.01,0.2640915107727051
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,4,1,power_law_1.01,0.2843091201782227
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,4,1,power_law_1.01,0.30392831802368164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,4,1,power_law_1.01,3.0859930419921873
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,4,1,power_law_1.01,0.22518720626831054
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,4,1,power_law_1.01,0.3644441604614258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,4,1,power_law_1.01,0.42876224517822265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,2,2,power_law_1.2,6.108342895507812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,4,1,power_law_1.01,0.5370988845825195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,4,1,power_law_1.01,0.6512633514404297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,4,1,power_law_1.01,0.9085395050048828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,4,1,power_law_1.01,1.1467014312744142
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,4,1,power_law_1.2,0.12516736030578612
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,4,1,power_law_1.2,0.1613849639892578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,4,1,power_law_1.01,1.6476460266113282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,4,1,power_law_1.2,0.2814835166931152
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,4,1,power_law_1.2,0.2973164749145508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,4,1,power_law_1.01,2.2212384033203127
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,4,1,power_law_1.2,0.30892032623291016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,4,1,power_law_1.2,0.3014803123474121
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,4,1,power_law_1.2,0.3384198379516602
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,4,1,power_law_1.2,0.32455039978027345
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,4,1,power_law_1.2,0.344642562866211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,4,1,power_law_1.2,0.3397126388549805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,4,1,power_law_1.2,0.39078655242919924
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,4,1,power_law_1.2,0.06238080024719238
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,4,1,power_law_1.2,0.09191360473632812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,4,1,power_law_1.2,0.37023361206054684
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,4,1,power_law_1.2,0.4010047912597656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,4,1,power_law_1.2,0.04815040111541748
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,4,1,power_law_1.2,0.4513024139404297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,4,1,power_law_1.2,0.5401894378662109
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,4,1,power_law_1.2,0.6342060852050782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,4,1,power_law_1.2,0.2601420783996582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,4,1,power_law_1.2,0.7874419403076172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,4,1,power_law_1.2,0.035253760814666746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,4,1,power_law_1.2,0.05089663982391357
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,4,1,power_law_1.2,0.9729824066162109
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,4,1,power_law_1.2,0.06888895988464355
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,4,1,power_law_1.2,0.09155455589294434
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,4,1,power_law_1.2,0.11369407653808594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,4,1,power_law_1.2,0.1705740737915039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,4,1,power_law_1.2,0.18053695678710938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,4,1,power_law_1.2,0.18608768463134767
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,4,1,power_law_1.2,0.19069631576538087
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,4,1,power_law_1.2,0.19021312713623045
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,4,1,power_law_1.2,0.19516544342041015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,4,1,power_law_1.2,0.35091968536376955
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,4,1,power_law_1.2,0.22786176681518552
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,4,1,power_law_1.2,0.23015743255615234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,4,1,power_law_1.2,0.23723775863647462
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,4,1,power_law_1.2,0.24585023880004883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,4,1,power_law_1.2,0.25167680740356446
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,4,1,power_law_1.2,0.2710432052612305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,4,1,power_law_1.2,1.3213369750976562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,4,1,power_law_1.2,0.2864473533630371
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,4,1,power_law_1.2,0.30631807327270505
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,4,1,power_law_1.2,0.37742080688476565
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,4,1,power_law_1.2,0.43352512359619144
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,4,1,power_law_1.2,0.5414630508422851
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,4,1,power_law_1.2,0.6663922882080078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,4,1,power_law_1.2,1.704442901611328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,1,4,balanced,0.08220928192138673
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,1,4,balanced,0.117640962600708
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,1,4,balanced,0.19714176177978515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,4,1,power_law_1.2,0.9199724578857422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,1,4,balanced,0.387248649597168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,1,4,balanced,0.748581085205078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,1,4,balanced,0.9108966064453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,1,4,balanced,1.4099801635742186
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,1,4,balanced,1.4207040405273437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,1,4,balanced,1.426183624267578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,1,4,balanced,1.4309312438964843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,1,4,balanced,1.4298060607910155
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,1,4,balanced,1.4424217224121094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,4,1,power_law_1.2,1.1616716766357422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,1,4,balanced,1.4691416931152343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,1,4,balanced,1.4960627746582031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,1,4,balanced,1.5216563415527342
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,1,4,balanced,1.559878387451172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,1,4,balanced,1.613744659423828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,1,4,balanced,1.6889715576171873
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,2,2,power_law_1.2,2.3013632202148435
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,1,4,balanced,1.7561651611328124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,1,4,balanced,2.0262208557128907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,1,4,balanced,2.136976623535156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,1,4,balanced,3.0202520751953124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,1,4,balanced,3.7830938720703124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,1,4,balanced,5.177354125976562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,1,4,balanced,0.06108287811279297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,1,4,balanced,0.08177791595458984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,4,1,power_law_1.2,2.3774745178222654
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,1,4,balanced,0.12999296188354492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,1,4,balanced,0.21149696350097655
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,1,4,balanced,0.38032257080078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,1,4,balanced,0.4781798553466797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,1,4,balanced,7.290396118164063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,1,4,balanced,0.5274137496948242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,1,4,balanced,0.5300556945800781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,1,4,balanced,0.5411225509643555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,1,4,balanced,0.5364953613281249
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,1,4,balanced,0.5276671981811523
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,1,4,balanced,0.5389030456542969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,1,4,balanced,0.5368576049804688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,1,4,balanced,0.5413119888305664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,1,4,balanced,0.5507430267333985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,1,4,balanced,0.5601382446289063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,1,4,balanced,0.5764620971679688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,1,4,balanced,0.6179264068603516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,1,4,balanced,0.6434496307373048
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,1,4,balanced,0.7145420837402343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,1,4,balanced,0.784421157836914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,1,4,balanced,1.1371353912353517
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,4,1,power_law_1.2,1.707678680419922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,1,4,balanced,10.840595703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,1,4,balanced,1.2528217315673829
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,1,4,power_law_1.01,0.1948838424682617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,1,4,balanced,1.7980120849609373
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,1,4,power_law_1.01,0.37413375854492187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,1,4,power_law_1.01,0.36803585052490234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,1,4,balanced,2.38967041015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,1,4,power_law_1.01,0.5258521652221679
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,1,4,power_law_1.01,0.5624870300292969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,1,4,balanced,3.1794509887695312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,1,4,power_law_1.01,0.7733554840087891
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,1,4,power_law_1.01,1.2604083251953124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,1,4,power_law_1.01,1.2249254608154296
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,1,4,balanced,4.209224853515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,1,4,power_law_1.01,1.2707584381103516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,1,4,power_law_1.01,1.32793212890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,1,4,balanced,14.366519775390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,1,4,power_law_1.01,1.3479808044433592
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,1,4,power_law_1.01,1.3745433044433593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,1,4,power_law_1.01,1.3962867736816407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,1,4,power_law_1.01,1.4314022827148438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,1,4,power_law_1.01,1.539612121582031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,1,4,power_law_1.01,1.604469757080078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,1,4,power_law_1.01,1.6645709228515624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,1,4,power_law_1.01,1.7280601501464843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,1,4,power_law_1.01,1.9878335571289063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,1,4,power_law_1.01,2.335074615478516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,1,4,power_law_1.01,2.6493234252929687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,1,4,power_law_1.01,3.806565246582031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,1,4,power_law_1.01,0.13503616333007812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,1,4,power_law_1.01,0.19370624542236328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,1,4,power_law_1.01,4.697172546386719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,1,4,power_law_1.01,0.17538816452026368
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,1,4,power_law_1.01,0.26799104690551756
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,1,4,power_law_1.01,0.313308162689209
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,1,4,power_law_1.01,0.39137409210205076
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,1,4,power_law_1.01,0.42245632171630854
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,1,4,power_law_1.01,6.4812646484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,1,4,power_law_1.01,0.4601113510131836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,1,4,power_law_1.01,0.4748979187011718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,1,4,power_law_1.01,0.4690329742431641
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,1,4,power_law_1.01,0.4780313491821289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,1,4,power_law_1.01,0.47502590179443355
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,1,4,power_law_1.01,0.4888780975341797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,1,4,power_law_1.01,0.48995326995849614
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,1,4,power_law_1.01,8.69161865234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,1,4,power_law_1.01,0.5092351913452149
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,1,4,power_law_1.01,0.5167103958129883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,1,4,power_law_1.01,0.550463981628418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,1,4,power_law_1.01,0.6081830215454102
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,1,4,power_law_1.01,0.6366336059570312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,1,4,power_law_1.01,0.7638041687011718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,4,1,power_law_1.2,3.0684909057617187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,1,4,power_law_1.01,0.875895004272461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,1,4,power_law_1.01,1.1953164672851562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,1,4,power_law_1.01,1.4785792541503906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,1,4,power_law_1.2,0.19969152450561523
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,1,4,power_law_1.2,0.25243263244628905
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,1,4,power_law_1.01,2.2220172119140624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,1,4,power_law_1.2,0.39104385375976564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,1,4,power_law_1.2,0.4566796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,1,4,power_law_1.01,2.8010482788085938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,1,4,power_law_1.2,0.5491571044921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,1,4,power_law_1.01,13.819407958984375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,1,4,power_law_1.2,0.8033663940429687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,1,4,power_law_1.01,3.975382995605469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,1,4,power_law_1.2,1.1369471740722656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,1,4,power_law_1.2,1.2516915130615236
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,1,4,power_law_1.2,1.2989683532714844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,1,4,power_law_1.2,1.3610073852539064
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,1,4,power_law_1.2,1.289740753173828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,1,4,power_law_1.01,5.247890014648438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,1,4,power_law_1.2,1.4406976318359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,1,4,power_law_1.2,1.4593664550781251
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,1,4,power_law_1.2,1.539199981689453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,1,4,power_law_1.2,1.5907878112792968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,1,4,power_law_1.2,1.607342071533203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,1,4,power_law_1.01,17.146063232421874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,4,1,power_law_1.2,2.208912048339844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,1,4,power_law_1.2,1.6807693481445312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,1,4,power_law_1.2,1.863625030517578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,1,4,power_law_1.2,2.085655059814453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,1,4,power_law_1.2,2.5394943237304686
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,1,4,power_law_1.2,0.12898431777954103
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,1,4,power_law_1.2,2.998712463378906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,1,4,power_law_1.2,0.15292672157287596
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,1,4,power_law_1.2,0.2047488021850586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,1,4,power_law_1.2,0.23263488769531251
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,1,4,power_law_1.2,0.3259891128540039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,1,4,power_law_1.2,0.3818329620361328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,1,4,power_law_1.2,4.291002807617188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,1,4,power_law_1.2,0.41198719024658204
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,1,4,power_law_1.2,0.4466777420043945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,1,4,power_law_1.2,0.48055809020996093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,1,4,power_law_1.2,0.4740748977661132
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,1,4,power_law_1.2,5.393721313476563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,1,4,power_law_1.2,0.4678732681274414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,1,4,power_law_1.2,0.47584255218505855
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,1,4,power_law_1.2,0.5011404800415039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,1,4,power_law_1.2,0.530516471862793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,1,4,power_law_1.2,0.5272473526000977
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,1,4,power_law_1.2,0.51936767578125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,1,4,power_law_1.2,6.522799072265625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,1,4,power_law_1.2,0.5669043350219727
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,1,4,power_law_1.2,0.6323507308959961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,1,4,power_law_1.2,0.6868313598632813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,1,4,power_law_1.2,0.80495361328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,1,4,power_law_1.2,0.9617100524902344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,1,4,power_law_1.2,1.235472640991211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,1,4,power_law_1.2,1.5854949951171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,2,2,balanced,0.0819711971282959
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,1,4,power_law_1.2,9.253778076171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,2,2,balanced,0.11660287857055665
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,2,2,balanced,0.19716735839843752
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,2,2,balanced,0.3866521453857422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,1,4,power_law_1.2,2.4399923706054687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,2,2,balanced,0.7396160125732422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,2,2,balanced,0.8989836883544922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,2,2,balanced,0.9006208038330078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,2,2,balanced,0.9113011169433595
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,2,2,balanced,0.9099890899658203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,2,2,balanced,1.3791949462890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,2,2,balanced,1.3923277282714843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,1,4,power_law_1.2,3.0057843017578127
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,2,2,balanced,1.4030502319335938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,2,2,balanced,1.4249253845214844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,1,4,power_law_1.2,5.58959228515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,2,2,balanced,1.4525479125976564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,2,2,balanced,1.4746022033691406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,2,2,balanced,1.5178854370117187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,2,2,balanced,1.5771725463867188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,2,2,balanced,1.6627494812011718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,2,2,balanced,1.738965759277344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,2,2,balanced,2.0171597290039065
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,1,4,power_law_1.2,16.37997802734375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,2,2,balanced,2.128143310546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,1,4,power_law_1.2,4.348606872558594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,2,2,balanced,2.960016784667969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,2,2,balanced,0.053625597953796386
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,2,2,balanced,0.07778687953948975
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,2,2,balanced,0.12973183631896973
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,2,2,balanced,3.77952880859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,2,2,balanced,0.2095961570739746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,2,2,balanced,0.3841279983520508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,2,2,balanced,0.48119167327880863
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,1,4,power_law_1.2,20.214490966796877
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,2,2,balanced,5.065430908203124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,2,2,balanced,0.4777638244628906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,2,2,balanced,0.47804927825927734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,2,2,balanced,0.481475830078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,2,2,balanced,0.5377561569213867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,2,2,balanced,0.5331468963623047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,2,2,balanced,0.5312409591674805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,2,2,balanced,0.5343884658813477
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,2,2,balanced,0.5405734252929688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,2,2,balanced,0.5486425781250001
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,2,2,balanced,0.5510067367553712
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,2,2,balanced,0.5608806228637695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,2,2,balanced,0.5912768173217773
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,2,2,balanced,0.6365836715698242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,2,2,balanced,0.6749887847900391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,2,2,balanced,7.008245849609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,2,2,balanced,0.747805404663086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,2,2,balanced,1.1214514923095702
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,2,2,balanced,1.2734272003173828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,2,2,power_law_1.01,0.11241344451904296
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,2,2,power_law_1.01,0.18189695358276367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,2,2,power_law_1.01,0.32271102905273436
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,2,2,balanced,1.7924237060546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,2,2,power_law_1.01,0.40236160278320315
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,2,2,power_law_1.01,0.5104896163940429
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,2,2,balanced,2.4639308166503904
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,2,2,power_law_1.01,0.7224076843261719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,2,2,power_law_1.01,0.7998745727539063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,2,2,balanced,9.871905517578124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,2,2,power_law_1.01,0.8214809417724609
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,2,2,balanced,3.318529357910156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,2,2,power_law_1.01,0.7908493041992187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,2,2,power_law_1.01,1.2921331787109376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,2,2,power_law_1.01,1.2839961242675781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,2,2,balanced,4.442668762207031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,2,2,power_law_1.01,1.3426829528808593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,2,2,power_law_1.01,1.2841075134277342
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,2,2,power_law_1.01,1.3782783508300782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,2,2,power_law_1.01,1.441361846923828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,2,2,power_law_1.01,1.519998779296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,2,2,power_law_1.01,1.5716390991210938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,2,2,power_law_1.01,1.6759846496582032
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,2,2,power_law_1.01,1.8180351257324219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,2,2,power_law_1.01,2.0915968322753904
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,2,2,power_law_1.01,2.4862629699707033
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,2,2,balanced,12.851226806640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,2,2,power_law_1.01,0.07808256149291992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,2,2,power_law_1.01,0.12457728385925293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,2,2,power_law_1.01,3.3096691894531247
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,2,2,power_law_1.01,0.13812607765197754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,2,2,power_law_1.01,0.19876224517822266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,2,2,power_law_1.01,0.272488956451416
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,2,2,power_law_1.01,0.3720064163208008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,2,2,power_law_1.01,0.3766003036499023
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,2,2,power_law_1.01,3.985028991699219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,2,2,power_law_1.01,0.3911974334716797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,2,2,power_law_1.01,0.4031078338623047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,2,2,power_law_1.01,0.445976333618164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,2,2,power_law_1.01,0.48182529449462896
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,2,2,power_law_1.01,0.46707839965820314
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,2,2,power_law_1.01,0.4726604843139649
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,2,2,power_law_1.01,0.495074577331543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,2,2,power_law_1.01,0.5090560150146485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,2,2,power_law_1.01,0.5205734252929688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,2,2,power_law_1.01,0.5360422515869141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,2,2,power_law_1.01,5.5802111816406255
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,2,2,power_law_1.01,0.5753894424438476
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,2,2,power_law_1.01,0.6362457656860352
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,2,2,power_law_1.01,0.7121753692626953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,2,2,power_law_1.01,0.8708262634277343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,2,2,power_law_1.01,1.2016307067871093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,2,2,power_law_1.01,1.3405401611328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,2,2,power_law_1.01,7.016807861328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,2,2,power_law_1.2,0.11133567810058594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,2,2,power_law_1.2,0.16119552612304688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,2,2,power_law_1.01,1.927262725830078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,2,2,power_law_1.2,0.3255116653442383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,2,2,power_law_1.2,0.3935795211791992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,2,2,power_law_1.01,2.4200460815429685
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,2,2,power_law_1.2,0.4650892639160157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,2,2,power_law_1.2,0.7666278076171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,2,2,power_law_1.2,0.792890853881836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,2,2,power_law_1.2,0.7720845031738282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,2,2,power_law_1.2,0.8377945709228516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,2,2,power_law_1.01,3.406252746582031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,2,2,power_law_1.2,1.283038787841797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,2,2,power_law_1.2,1.3376460266113281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,2,2,power_law_1.2,1.3613516235351564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,2,2,power_law_1.01,11.3387109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,2,2,power_law_1.2,1.362540740966797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,2,2,power_law_1.2,1.4139059448242188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,2,2,power_law_1.01,4.745329895019531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,2,2,power_law_1.2,1.4659263610839843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,2,2,power_law_1.2,1.5364186096191408
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,2,2,power_law_1.2,1.6178675842285155
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,2,2,power_law_1.2,1.7364031982421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,2,2,power_law_1.2,1.8711283874511717
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,2,2,power_law_1.2,2.184340515136719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,2,2,power_law_1.01,14.014555664062499
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,2,2,power_law_1.2,2.6385638427734373
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,2,2,power_law_1.2,0.07504000186920165
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,2,2,power_law_1.2,0.10247296333312987
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,2,2,power_law_1.2,0.13920767784118654
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,2,2,power_law_1.2,0.18073087692260742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,2,2,power_law_1.2,0.24462848663330078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,2,2,power_law_1.2,3.5436517333984376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,2,2,power_law_1.2,0.3701363372802734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,2,2,power_law_1.2,0.3710079956054687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,2,2,power_law_1.2,0.4288934326171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,2,2,power_law_1.2,4.139939880371093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,2,2,power_law_1.2,0.411383056640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,2,2,power_law_1.2,0.4499635314941406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,2,2,power_law_1.2,0.45950462341308596
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,2,2,power_law_1.2,0.5021043014526367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,2,2,power_law_1.2,0.49755775451660156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,2,2,power_law_1.2,0.503636474609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,2,2,power_law_1.2,0.5140428924560547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,2,2,power_law_1.2,0.5192793655395508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,2,2,power_law_1.2,0.5629811096191406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,2,2,power_law_1.2,0.5962393569946289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,2,2,power_law_1.2,0.6527385711669922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,2,2,power_law_1.2,5.899824829101563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,2,2,power_law_1.2,0.7431180572509766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,2,2,power_law_1.2,0.8982220458984376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,2,2,power_law_1.2,1.162656021118164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,2,2,power_law_1.2,7.197259521484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,2,2,power_law_1.2,1.3885133361816406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,4,1,balanced,0.08679167747497558
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,4,1,balanced,0.12365311622619628
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,4,1,balanced,0.20291584014892577
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,2,2,power_law_1.2,1.9964927673339843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,4,1,balanced,0.39507518768310546
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,2,2,power_law_1.2,2.5557926940917968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,4,1,balanced,0.7453580474853515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,4,1,balanced,0.907360610961914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,4,1,balanced,0.9059916687011718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,4,1,balanced,0.9176473236083984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,4,1,balanced,0.9243564605712891
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,4,1,balanced,0.9247821044921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,4,1,balanced,0.9400953674316407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,2,2,power_law_1.2,3.497313232421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,4,1,balanced,0.9547103881835938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,4,1,balanced,1.3977964782714845
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,4,1,balanced,1.422967071533203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,4,1,balanced,1.4455769348144532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,2,2,power_law_1.2,11.7268408203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,4,1,balanced,1.4816082763671876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,4,1,balanced,1.5380972290039063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,4,1,balanced,1.6172186279296874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,4,1,balanced,1.6950930786132812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,4,1,balanced,2.1379052734375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,2,2,power_law_1.2,15.063077392578126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,4,1,balanced,2.1937190246582032
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,4,1,balanced,0.06607679843902588
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,4,1,balanced,0.08750144004821778
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,4,1,balanced,2.9499755859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,4,1,balanced,0.13878016471862792
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,4,1,balanced,0.22773311614990233
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,4,1,balanced,0.40112640380859377
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,4,1,balanced,3.7965402221679687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,4,1,balanced,0.5023001480102539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,4,1,balanced,0.5054470443725586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,4,1,balanced,0.5255007934570313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,4,1,balanced,0.5142995071411133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,4,1,balanced,0.5127539062500001
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,4,1,balanced,0.5190713500976563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,4,1,balanced,0.5246412658691406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,4,1,balanced,0.5914694213867188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,4,1,balanced,5.254429931640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,4,1,balanced,0.5910534286499024
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,4,1,balanced,0.5984454345703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,4,1,balanced,0.6092825698852539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,4,1,balanced,0.6243628692626954
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,4,1,balanced,0.6340044784545898
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,4,1,balanced,0.6615763092041016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,4,1,balanced,0.7254335784912109
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,4,1,balanced,0.7931660461425781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,4,1,balanced,6.736586303710938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,4,1,balanced,1.2174425506591797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,4,1,balanced,1.3682765197753906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,4,1,power_law_1.01,0.08781375885009765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,4,1,power_law_1.01,0.12072704315185548
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,4,1,balanced,1.990832061767578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,4,1,power_law_1.01,0.20241920471191407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,4,1,power_law_1.01,0.37006080627441407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,4,1,balanced,2.5570886230468752
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,4,1,power_law_1.01,0.445910415649414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,4,1,power_law_1.01,0.7226329803466797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,4,1,balanced,9.587158813476563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,4,1,power_law_1.01,0.7792025756835937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,4,1,power_law_1.01,0.8129446411132812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,4,1,balanced,3.3314239501953127
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,4,1,power_law_1.01,0.840860824584961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,4,1,power_law_1.01,0.8427724456787109
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,4,1,power_law_1.01,0.8913740539550782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,4,1,power_law_1.01,0.9183999633789062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,4,1,balanced,4.44524658203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,4,1,power_law_1.01,1.2994566345214844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,4,1,power_law_1.01,1.377098846435547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,4,1,power_law_1.01,1.4107936096191405
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,4,1,power_law_1.01,1.4282412719726563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,4,1,balanced,12.531976318359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,4,1,power_law_1.01,1.515032958984375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,4,1,power_law_1.01,1.6474464416503907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,4,1,power_law_1.01,1.7400921630859376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,4,1,power_law_1.01,1.9775520324707032
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,4,1,power_law_1.01,2.319718475341797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,2,2,power_law_1.2,4.511980895996094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,4,1,power_law_1.01,0.06376832008361817
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,4,1,power_law_1.01,2.935553283691406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,4,1,power_law_1.01,0.09025152206420899
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,4,1,power_law_1.01,0.13979583740234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,4,1,power_law_1.01,0.2026188850402832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,4,1,power_law_1.01,0.2561644744873047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,4,1,power_law_1.01,3.6866906738281253
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,4,1,power_law_1.01,0.39295425415039065
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,4,1,power_law_1.01,0.4202239990234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,4,1,power_law_1.01,0.43817600250244143
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,4,1,power_law_1.01,0.468963851928711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,4,1,power_law_1.01,0.4577318572998047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,4,1,power_law_1.01,0.47116928100585936
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,4,1,power_law_1.01,0.4809875106811523
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,4,1,power_law_1.01,0.5373907089233398
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,4,1,power_law_1.01,5.098308410644531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,4,1,power_law_1.01,0.5529836654663086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,4,1,power_law_1.01,0.5568499374389648
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,4,1,power_law_1.01,0.5571456146240235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,4,1,power_law_1.01,0.5866572952270508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,4,1,power_law_1.01,0.6437702178955078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,4,1,power_law_1.01,0.6902483367919923
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,4,1,power_law_1.01,0.7767916870117187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,4,1,power_law_1.01,0.9241645050048828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,4,1,power_law_1.01,1.1675641632080078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,4,1,power_law_1.01,6.729593505859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,4,1,power_law_1.01,1.4162567138671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,4,1,power_law_1.2,0.08790528297424316
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,4,1,power_law_1.2,0.12256064414978027
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,4,1,power_law_1.01,1.9870930480957032
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,4,1,power_law_1.2,0.1985696029663086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,4,1,power_law_1.2,0.3533472061157227
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,4,1,power_law_1.2,0.4185459136962891
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,4,1,power_law_1.01,2.414300231933594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,4,1,power_law_1.2,0.7082828521728516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,4,1,power_law_1.2,0.7688134765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,4,1,power_law_1.2,0.8198086547851563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,4,1,power_law_1.2,0.8436441802978516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,4,1,power_law_1.2,0.865283203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,4,1,power_law_1.01,3.4923187255859376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,4,1,power_law_1.2,0.9161004638671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,4,1,power_law_1.01,9.490726928710938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,4,1,power_law_1.2,0.9461401367187501
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,4,1,power_law_1.2,1.358321990966797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,4,1,power_law_1.2,1.3931770324707031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,4,1,power_law_1.01,4.485472106933594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,4,1,power_law_1.2,1.4796493530273438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,4,1,power_law_1.2,1.5119065856933593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,4,1,power_law_1.2,1.5457286071777343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,4,1,power_law_1.2,1.6626124572753906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,4,1,power_law_1.2,1.7535116577148435
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,4,1,power_law_1.01,12.28976806640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,4,1,power_law_1.2,2.0554841613769534
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,4,1,power_law_1.2,2.3809625244140626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,4,1,power_law_1.2,0.06446720123291015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,4,1,power_law_1.2,0.08800895690917969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,4,1,power_law_1.2,0.1397599983215332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,4,1,power_law_1.2,3.069691467285156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,4,1,power_law_1.2,0.17635583877563477
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,4,1,power_law_1.2,0.23659072875976564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,4,1,power_law_1.2,0.3989958572387695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,4,1,power_law_1.2,0.41514049530029296
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,4,1,power_law_1.2,3.695816345214844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,4,1,power_law_1.2,0.4416838455200195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,4,1,power_law_1.2,0.45767105102539063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,4,1,power_law_1.2,0.46691326141357425
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,4,1,power_law_1.2,0.4813568115234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,4,1,power_law_1.2,0.4866912078857422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,4,1,power_law_1.2,0.5566188812255859
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,4,1,power_law_1.2,0.5569631958007812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,4,1,power_law_1.2,0.5824556732177735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,4,1,power_law_1.2,5.110649108886719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,4,1,power_law_1.2,0.5937503814697266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,4,1,power_law_1.2,0.6158188629150391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,4,1,power_law_1.2,0.6534361267089844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,4,1,power_law_1.2,0.7085343933105468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,4,1,power_law_1.2,0.8606924438476563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,4,1,power_law_1.2,0.9090694427490235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,4,1,power_law_1.2,1.213736343383789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,4,1,power_law_1.2,6.557640991210937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,4,1,power_law_1.2,1.4388908386230468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,1,4,balanced,0.09157888412475586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,1,4,balanced,0.11206912040710448
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,1,4,balanced,0.2040793609619141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,4,1,power_law_1.2,1.9728294372558595
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,1,4,balanced,0.4224972915649414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,1,4,balanced,0.7042368316650391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,4,1,power_law_1.2,2.496135711669922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,1,4,balanced,1.3356825256347657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,1,4,balanced,1.9809817504882812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,1,4,balanced,1.9894041442871093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,1,4,balanced,1.9992883300781252
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,1,4,balanced,2.0037324523925784
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,1,4,balanced,3.0490737915039063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,1,4,balanced,3.0640551757812498
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,1,4,balanced,3.065035400390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,1,4,balanced,3.094468994140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,4,1,power_law_1.2,3.5231640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,4,1,power_law_1.2,9.589398193359376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,1,4,balanced,3.113816223144531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,1,4,balanced,3.127069396972656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,1,4,balanced,3.19602294921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,1,4,balanced,3.2784359741210936
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,1,4,balanced,3.4155532836914064
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,1,4,balanced,3.6279922485351563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,1,4,balanced,3.8408843994140627
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,1,4,balanced,4.147155151367188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,1,4,balanced,0.053752322196960446
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,1,4,balanced,4.7208972167968755
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,1,4,balanced,0.07077888011932373
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,1,4,balanced,0.11897600173950196
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,1,4,balanced,0.2140262413024902
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,1,4,balanced,5.189893188476562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,1,4,balanced,0.36971775054931644
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,1,4,balanced,0.6819404602050781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,1,4,balanced,0.9981708526611328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,1,4,balanced,1.000348129272461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,4,1,power_law_1.2,4.539718322753906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,1,4,balanced,1.0041343688964843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,1,4,balanced,7.1103973388671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,1,4,balanced,1.011278076171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,1,4,balanced,1.081085433959961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,1,4,balanced,1.0701286315917968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,1,4,balanced,1.0811199951171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,1,4,balanced,1.1275392150878907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,1,4,balanced,1.132085723876953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,4,1,power_law_1.2,12.61669921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,1,4,balanced,1.120385284423828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,1,4,balanced,1.130449905395508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,1,4,balanced,1.1915878295898437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,1,4,balanced,1.2361971282958986
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,1,4,balanced,9.982543334960937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,1,4,balanced,1.3434176635742188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,1,4,balanced,1.3991410827636719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,1,4,balanced,1.5081471252441407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,1,4,balanced,1.6666099548339843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,1,4,power_law_1.01,0.20444927215576172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,1,4,power_law_1.01,0.42846847534179683
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,1,4,balanced,2.0010662841796876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,1,4,power_law_1.01,0.7016729736328126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,1,4,power_law_1.01,0.7068134307861328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,1,4,balanced,2.7748184204101562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,1,4,power_law_1.01,0.9149171447753905
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,1,4,balanced,3.363222961425781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,1,4,power_law_1.01,1.2415270233154296
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,1,4,power_law_1.01,1.2257843017578125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,1,4,power_law_1.01,1.715778503417969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,1,4,power_law_1.01,1.5631936645507811
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,1,4,power_law_1.01,1.5435801696777343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,1,4,balanced,4.522497253417969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,1,4,balanced,14.446016845703124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,1,4,power_law_1.01,2.3580032348632813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,1,4,power_law_1.01,2.4517811584472655
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,1,4,power_law_1.01,2.3959078979492188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,1,4,power_law_1.01,2.5155366516113284
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,1,4,power_law_1.01,2.6080538940429685
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,1,4,power_law_1.01,2.7840676879882813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,1,4,power_law_1.01,2.8648666381835937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,1,4,power_law_1.01,2.9757479858398437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,1,4,power_law_1.01,3.1000103759765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,1,4,power_law_1.01,3.523531494140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,1,4,power_law_1.01,3.7928512573242186
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,1,4,power_law_1.01,0.11761280059814454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,1,4,power_law_1.01,0.2121881675720215
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,1,4,power_law_1.01,0.3722982406616211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,1,4,power_law_1.01,4.440195922851562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,1,4,power_law_1.01,0.4105126571655274
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,1,4,power_law_1.01,0.5258675384521484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,1,4,power_law_1.01,0.5751897430419921
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,1,4,power_law_1.01,5.292634887695312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,1,4,power_law_1.01,0.7028288269042969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,1,4,power_law_1.01,0.7854732513427735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,1,4,power_law_1.01,0.7883264160156249
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,1,4,power_law_1.01,0.774603500366211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,1,4,power_law_1.01,0.8697586822509766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,1,4,power_law_1.01,7.224414672851563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,1,4,power_law_1.01,0.8771212768554687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,1,4,power_law_1.01,0.9339199829101563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,1,4,power_law_1.01,0.8932543945312499
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,1,4,power_law_1.01,0.8513651275634766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,1,4,power_law_1.01,0.9815744018554687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,1,4,power_law_1.01,8.65804443359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,1,4,power_law_1.01,1.0059187316894531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,1,4,power_law_1.01,1.0915494537353516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,1,4,power_law_1.01,1.1202098846435546
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,1,4,power_law_1.01,1.1556902313232422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,1,4,power_law_1.01,1.3505267333984374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,1,4,power_law_1.01,1.6021209716796876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,1,4,power_law_1.01,1.8683135986328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,1,4,power_law_1.01,2.3419815063476563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,1,4,power_law_1.2,0.20706560134887697
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,1,4,power_law_1.2,0.4225932693481445
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,1,4,power_law_1.01,12.187294921874999
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,1,4,power_law_1.2,0.5539827346801758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,1,4,power_law_1.2,0.6988787078857422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,1,4,power_law_1.2,0.8990860748291016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,1,4,power_law_1.2,1.1009574127197266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,1,4,power_law_1.2,1.0649842834472656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,1,4,power_law_1.01,2.9742669677734375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,1,4,power_law_1.01,4.106697082519531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,1,4,power_law_1.2,1.3780096435546876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,1,4,power_law_1.2,1.4319039916992187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,1,4,power_law_1.2,1.4788517761230469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,1,4,power_law_1.01,14.920073242187499
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,1,4,power_law_1.01,5.4322534179687505
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,1,4,power_law_1.2,2.196746215820313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,1,4,power_law_1.2,2.491663360595703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,1,4,power_law_1.2,2.489268493652344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,1,4,power_law_1.2,2.5311065673828126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,1,4,power_law_1.2,2.5629055786132815
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,1,4,power_law_1.2,2.572062683105469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,1,4,power_law_1.2,2.7744677734375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,1,4,power_law_1.2,3.0920986938476562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,1,4,power_law_1.2,3.2943872070312503
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,1,4,power_law_1.2,3.805592346191406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,1,4,power_law_1.2,0.11756287574768068
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,1,4,power_law_1.2,4.118434448242188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,1,4,power_law_1.2,0.21144447326660157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,1,4,power_law_1.2,4.7510833740234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,1,4,power_law_1.2,0.3003891181945801
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,1,4,power_law_1.2,0.39310848236083984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,1,4,power_law_1.2,0.4688127899169922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,1,4,power_law_1.2,0.5692736053466797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,1,4,power_law_1.2,5.641327514648437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,1,4,power_law_1.2,0.6546125030517578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,1,4,power_law_1.2,0.7996774291992187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,1,4,power_law_1.2,0.8000870513916016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,1,4,power_law_1.2,0.7493734741210938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,1,4,power_law_1.2,0.8712319946289062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,1,4,power_law_1.2,7.463225708007812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,1,4,power_law_1.2,0.8923661041259766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,1,4,power_law_1.2,0.8916019439697266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,1,4,power_law_1.2,0.9221298980712891
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,1,4,power_law_1.2,0.9134732818603515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,1,4,power_law_1.2,0.9002022552490235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,1,4,power_law_1.2,0.9445452880859374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,1,4,power_law_1.2,1.0614412689208985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,1,4,power_law_1.2,1.1968486022949218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,1,4,power_law_1.2,1.2222758483886718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,1,4,power_law_1.2,1.4115711975097658
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,1,4,power_law_1.2,10.473973388671876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,1,4,power_law_1.2,1.7257101440429685
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,2,2,balanced,0.09172736167907715
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,1,4,power_law_1.2,2.026497344970703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,2,2,balanced,0.12492416381835938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,2,2,balanced,0.19743871688842773
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,2,2,balanced,0.4082854461669922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,1,4,power_law_1.2,2.736971435546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,2,2,balanced,0.6901055908203124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,1,4,power_law_1.2,3.1329025268554687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,2,2,balanced,1.3287014770507812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,2,2,balanced,1.9778419494628907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,2,2,balanced,1.9861016845703126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,1,4,power_law_1.2,4.282272033691406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,1,4,power_law_1.2,14.248323974609374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,2,2,balanced,1.9850636291503907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,2,2,balanced,1.995863037109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,1,4,power_law_1.2,5.404646606445313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,2,2,balanced,2.009676818847656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,2,2,balanced,2.021017608642578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,2,2,balanced,2.0230221557617187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,1,4,power_law_1.2,16.54892333984375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,2,2,balanced,3.0100326538085938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,2,2,balanced,3.029401550292969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,2,2,balanced,3.0506878662109376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,2,2,balanced,3.1096038818359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,2,2,balanced,3.193861083984375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,2,2,balanced,3.316405639648438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,2,2,balanced,3.549224853515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,2,2,balanced,0.05299712181091308
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,2,2,balanced,3.7962573242187503
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,2,2,balanced,0.07158912181854248
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,2,2,balanced,0.12080127716064452
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,2,2,balanced,0.21073919296264648
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,2,2,balanced,4.105647277832031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,2,2,balanced,0.3689228820800781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,2,2,balanced,0.6798963165283203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,2,2,balanced,4.721164855957031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,2,2,balanced,0.9998464202880859
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,2,2,balanced,5.487031249999999
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,2,2,balanced,1.0013900756835938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,2,2,balanced,1.0054144287109374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,2,2,balanced,1.010324478149414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,2,2,balanced,1.0231488037109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,2,2,balanced,1.029260787963867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,2,2,balanced,1.1294271850585937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,2,2,balanced,7.075838623046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,2,2,balanced,1.1754201507568358
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,2,2,balanced,1.032643814086914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,2,2,balanced,1.161032943725586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,2,2,balanced,1.1537267303466796
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,2,2,balanced,1.2080754852294922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,2,2,balanced,1.2806874084472657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,2,2,balanced,1.2731558227539064
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,2,2,balanced,1.3815539550781248
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,2,2,balanced,9.345286254882812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,2,2,balanced,1.519569854736328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,2,2,power_law_1.01,0.12834815979003905
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,2,2,balanced,1.6434240722656248
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,2,2,power_law_1.01,0.18825471878051758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,2,2,balanced,1.9855448913574218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,2,2,power_law_1.01,0.4087526321411133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,2,2,power_law_1.01,0.514648323059082
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,2,2,power_law_1.01,0.7227033233642578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,2,2,balanced,2.7618212890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,2,2,power_law_1.01,1.0113574218749999
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,2,2,power_law_1.01,1.0924262237548827
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,2,2,balanced,3.347819519042969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,2,2,power_law_1.01,1.449969940185547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,2,2,power_law_1.01,1.4342080688476562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,2,2,power_law_1.01,1.48276611328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,2,2,balanced,4.39886474609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,2,2,balanced,13.43853271484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,2,2,power_law_1.01,1.5127757263183592
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,2,2,power_law_1.01,1.5976524353027344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,2,2,power_law_1.01,1.5989939880371093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,2,2,power_law_1.01,2.451110382080078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,2,2,power_law_1.01,2.5261209106445315
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,2,2,power_law_1.01,2.5702655029296873
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,2,2,power_law_1.01,2.668872985839844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,2,2,power_law_1.01,2.9237350463867187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,2,2,power_law_1.01,3.043291015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,2,2,power_law_1.01,3.457015075683594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,2,2,power_law_1.01,3.7011993408203123
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,2,2,power_law_1.01,0.07150335788726807
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,2,2,power_law_1.01,0.1204032039642334
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,2,2,power_law_1.01,0.2107622337341309
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,2,2,power_law_1.01,0.28844800949096683
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,2,2,power_law_1.01,0.39792255401611326
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,2,2,power_law_1.01,4.289240417480469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,2,2,power_law_1.01,0.5258598327636719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,2,2,power_law_1.01,0.5890316772460937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,2,2,power_law_1.01,4.980707702636719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,2,2,power_law_1.01,0.7417523193359374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,2,2,power_law_1.01,0.7391731262207031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,2,2,power_law_1.01,0.7591001892089844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,2,2,power_law_1.01,0.768264923095703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,2,2,power_law_1.01,6.1799554443359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,2,2,power_law_1.01,0.7994713592529297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,2,2,power_law_1.01,0.812390365600586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,2,2,power_law_1.01,0.9010227203369141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,2,2,power_law_1.01,0.9549734497070312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,2,2,power_law_1.01,0.9628620910644532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,2,2,power_law_1.01,0.9899276733398438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,2,2,power_law_1.01,7.682955322265625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,2,2,power_law_1.01,1.0519475555419922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,2,2,power_law_1.01,1.1055526733398438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,2,2,power_law_1.01,1.1752973175048829
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,2,2,power_law_1.01,1.3096946716308593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,2,2,power_law_1.01,1.590502471923828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,2,2,power_law_1.2,0.11689599990844726
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,2,2,power_law_1.2,0.18537216186523436
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,2,2,power_law_1.01,2.3992985534667968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,2,2,power_law_1.01,10.666085205078124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,2,2,power_law_1.2,0.32008575439453124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,2,2,power_law_1.01,1.9213221740722655
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,2,2,power_law_1.2,0.48782718658447266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,2,2,power_law_1.01,2.874805908203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,2,2,power_law_1.2,0.6804851531982422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,2,2,power_law_1.2,0.8806259155273437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,2,2,power_law_1.2,1.0621542358398437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,2,2,power_law_1.01,3.828597717285156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,2,2,power_law_1.2,1.3819801330566406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,2,2,power_law_1.2,1.471199951171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,2,2,power_law_1.2,1.4358297729492189
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,2,2,power_law_1.2,1.4780787658691406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,2,2,power_law_1.01,13.765003662109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,2,2,power_law_1.01,5.063473815917969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,2,2,power_law_1.2,1.5792524719238281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,2,2,power_law_1.2,1.6033485412597657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,2,2,power_law_1.2,2.487466278076172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,2,2,power_law_1.2,2.48094970703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,2,2,power_law_1.2,2.5666290283203126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,2,2,power_law_1.2,2.7608178710937503
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,2,2,power_law_1.2,3.000715637207031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,2,2,power_law_1.2,3.2016537475585936
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,2,2,power_law_1.2,3.5459213256835938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,2,2,power_law_1.2,3.865830383300781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,2,2,power_law_1.2,0.07153664112091065
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,2,2,power_law_1.2,0.1191104030609131
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,2,2,power_law_1.2,0.18431488037109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,2,2,power_law_1.2,0.280263671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,2,2,power_law_1.2,4.38780029296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,2,2,power_law_1.2,0.3828838348388672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,2,2,power_law_1.2,0.4679999923706054
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,2,2,power_law_1.2,0.5466265487670898
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,2,2,power_law_1.2,5.1846435546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,2,2,power_law_1.2,0.697991714477539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,2,2,power_law_1.2,0.749164810180664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,2,2,power_law_1.2,0.7296051025390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,2,2,power_law_1.2,6.566268310546874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,2,2,power_law_1.2,0.7826265716552735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,2,2,power_law_1.2,0.7878208160400391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,2,2,power_law_1.2,0.8089292907714845
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,2,2,power_law_1.2,0.9380915069580078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,2,2,power_law_1.2,0.9466547393798829
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,2,2,power_law_1.2,1.0129958343505858
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,2,2,power_law_1.2,8.21017333984375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,2,2,power_law_1.2,0.9829606628417968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,2,2,power_law_1.2,1.0902655792236327
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,2,2,power_law_1.2,1.1893235015869141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,2,2,power_law_1.2,1.2940876770019532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,2,2,power_law_1.2,1.3891290283203124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,2,2,power_law_1.2,1.6183935546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,2,2,power_law_1.2,1.9182989501953123
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,4,1,balanced,0.09478848457336427
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,4,1,balanced,0.12389568328857421
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,4,1,balanced,0.1939583969116211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,2,2,power_law_1.2,2.627613525390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,4,1,balanced,0.41806655883789057
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,4,1,balanced,0.7002105712890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,2,2,power_law_1.2,10.903712158203124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,2,2,power_law_1.2,3.0393304443359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,4,1,balanced,1.3407283020019531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,2,2,power_law_1.2,3.9519232177734374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,4,1,balanced,1.9855949401855468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,2,2,power_law_1.2,14.473526611328126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,4,1,balanced,1.9901011657714842
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,4,1,balanced,2.0011744689941406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,2,2,power_law_1.2,5.289830322265625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,4,1,balanced,2.0049171447753906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,4,1,balanced,2.0206739807128904
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,4,1,balanced,2.023079681396484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,4,1,balanced,2.041777954101563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,4,1,balanced,2.065021514892578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,4,1,balanced,2.0835040283203123
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,4,1,balanced,2.105199432373047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,4,1,balanced,3.0802783203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,4,1,balanced,3.277840576171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,4,1,balanced,3.1497982788085936
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,4,1,balanced,3.4654867553710935
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,4,1,balanced,0.06527488231658936
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,4,1,balanced,3.7215634155273434
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,4,1,balanced,0.08699199676513672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,4,1,balanced,0.1356563186645508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,4,1,balanced,0.227270393371582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,4,1,balanced,4.11430908203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,4,1,balanced,0.3942111968994141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,4,1,balanced,0.7381382751464843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,4,1,balanced,5.317172241210938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,4,1,balanced,1.0980652618408204
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,4,1,balanced,5.642758178710937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,4,1,balanced,1.0942982482910157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,4,1,balanced,1.0941548919677735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,4,1,balanced,1.101502685546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,4,1,balanced,1.1088703918457032
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,4,1,balanced,1.1383673858642578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,4,1,balanced,1.1292896270751953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,4,1,balanced,1.1332403564453126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,4,1,balanced,1.1480006408691406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,4,1,balanced,1.1592915344238282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,4,1,balanced,7.165941772460937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,4,1,balanced,1.3647027587890626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,4,1,balanced,1.3777049255371092
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,4,1,balanced,1.3328115844726562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,4,1,balanced,1.4509184265136719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,4,1,balanced,1.5166143798828124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,4,1,balanced,1.6890707397460936
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,4,1,balanced,1.7875462341308594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,4,1,power_law_1.01,0.0916761589050293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,4,1,power_law_1.01,0.12165439605712891
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,4,1,power_law_1.01,0.19148288726806642
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,4,1,balanced,9.546737670898438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,4,1,balanced,2.0515104675292966
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,4,1,power_law_1.01,0.41674945831298826
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,4,1,power_law_1.01,0.6672672271728516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,4,1,balanced,3.0736871337890626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,4,1,power_law_1.01,0.9195200347900391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,4,1,power_law_1.01,1.0464614105224608
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,4,1,balanced,3.59375244140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,4,1,power_law_1.01,1.3464218139648438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,4,1,power_law_1.01,1.4268608093261719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,4,1,power_law_1.01,1.4892153930664063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,4,1,power_law_1.01,1.5386778259277345
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,4,1,power_law_1.01,1.5548492431640626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,4,1,balanced,4.920423583984375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,4,1,power_law_1.01,1.6310336303710937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,4,1,power_law_1.01,1.719768371582031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,4,1,balanced,14.17771484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,4,1,power_law_1.01,1.8000038146972657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,4,1,power_law_1.01,1.9239712524414063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,4,1,power_law_1.01,2.5904031372070313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,4,1,power_law_1.01,2.797881469726563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,4,1,power_law_1.01,2.9722271728515626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,4,1,power_law_1.01,3.317068786621094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,4,1,power_law_1.01,3.6262957763671877
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,4,1,power_law_1.01,0.06598527908325195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,4,1,power_law_1.01,4.244922180175782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,4,1,power_law_1.01,0.08655360221862793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,4,1,power_law_1.01,0.13572928428649902
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,4,1,power_law_1.01,0.23120960235595703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,4,1,power_law_1.01,4.920838928222656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,4,1,power_law_1.01,0.3695481491088867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,4,1,power_law_1.01,0.5106508636474609
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,4,1,power_law_1.01,0.5864191818237304
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,4,1,power_law_1.01,0.7536665344238281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,4,1,power_law_1.01,0.7795398712158204
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,4,1,power_law_1.01,6.118119506835937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,4,1,power_law_1.01,0.8204402923583984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,4,1,power_law_1.01,0.8473414611816406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,4,1,power_law_1.01,0.8642835235595703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,4,1,power_law_1.01,0.8923379516601562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,4,1,power_law_1.01,0.9241011047363281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,4,1,power_law_1.01,0.9349849700927735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,4,1,power_law_1.01,0.9712895965576172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,4,1,power_law_1.01,7.384301147460937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,4,1,power_law_1.01,1.107639694213867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,4,1,power_law_1.01,1.1620294189453124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,4,1,power_law_1.01,1.235261459350586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,4,1,power_law_1.01,1.3452780151367187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,4,1,power_law_1.01,1.4678924560546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,4,1,power_law_1.01,1.7110322570800782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,4,1,power_law_1.01,1.9755014038085938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,4,1,power_law_1.2,0.09417023658752441
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,4,1,power_law_1.2,0.1217081642150879
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,4,1,power_law_1.01,2.497941131591797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,4,1,power_law_1.2,0.18979135513305664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,4,1,power_law_1.01,10.272608642578124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,4,1,power_law_1.2,0.41803966522216796
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,4,1,power_law_1.01,2.9982452392578125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,4,1,power_law_1.2,0.6344102478027344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,4,1,power_law_1.2,0.8289836883544922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,4,1,power_law_1.2,0.9786675262451172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,4,1,power_law_1.01,4.043499450683593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,4,1,power_law_1.2,1.3135360717773437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,4,1,power_law_1.2,1.3833279418945312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,4,1,power_law_1.2,1.437328643798828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,4,1,power_law_1.2,1.483019561767578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,4,1,power_law_1.2,1.5503654479980469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,4,1,power_law_1.01,12.853695068359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,4,1,power_law_1.01,5.065078430175781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,4,1,power_law_1.2,1.6326751708984375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,4,1,power_law_1.2,1.711231384277344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,4,1,power_law_1.2,1.830140838623047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,4,1,power_law_1.2,1.9360409545898438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,4,1,power_law_1.2,2.6254727172851564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,4,1,power_law_1.2,2.8550399780273437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,4,1,power_law_1.2,3.0285491943359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,4,1,power_law_1.2,3.407395935058594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,4,1,power_law_1.2,0.06805823802947998
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,4,1,power_law_1.2,3.791285095214844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,4,1,power_law_1.2,0.08649408340454101
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,4,1,power_law_1.2,4.339511108398438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,4,1,power_law_1.2,0.13706303596496583
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,4,1,power_law_1.2,0.2279155158996582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,4,1,power_law_1.2,0.35358657836914065
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,4,1,power_law_1.2,0.4640044784545898
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,4,1,power_law_1.2,5.007512817382812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,4,1,power_law_1.2,0.5465644836425781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,4,1,power_law_1.2,0.7325113677978516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,4,1,power_law_1.2,0.7617619323730469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,4,1,power_law_1.2,0.7897433471679688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,4,1,power_law_1.2,6.401875610351563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,4,1,power_law_1.2,0.8169548797607422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,4,1,power_law_1.2,0.8575296020507812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,4,1,power_law_1.2,0.8805299377441406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,4,1,power_law_1.2,0.9121190643310546
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,4,1,power_law_1.2,0.9355967712402343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,4,1,power_law_1.2,0.9583347320556641
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,4,1,power_law_1.2,7.565765380859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,4,1,power_law_1.2,1.1174291229248046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,4,1,power_law_1.2,1.198062744140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,4,1,power_law_1.2,1.2696966552734374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,4,1,power_law_1.2,1.4020384216308595
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,4,1,power_law_1.2,1.5176416015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,4,1,power_law_1.2,1.774703369140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,1,4,balanced,0.047884798049926756
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,4,1,power_law_1.2,2.0616487121582034
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,1,4,balanced,0.0800819206237793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,1,4,balanced,0.12545023918151857
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,4,1,power_law_1.2,10.304288330078126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,1,4,balanced,0.22969600677490237
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,4,1,power_law_1.2,2.5613754272460936
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,1,4,balanced,0.05186560153961182
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,1,4,balanced,0.4600716781616211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,1,4,balanced,0.4421030426025391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,1,4,balanced,0.44664958953857425
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,4,1,power_law_1.2,3.1067956542968753
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,1,4,balanced,0.6495014190673828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,1,4,balanced,0.6546329498291016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,1,4,balanced,0.6665561676025391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,1,4,balanced,0.6677632141113281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,1,4,balanced,0.675937271118164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,1,4,balanced,0.6986022186279297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,1,4,balanced,0.7128102111816406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,1,4,balanced,0.7300697326660156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,4,1,power_law_1.2,4.167127685546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,1,4,balanced,0.7782643127441407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,1,4,balanced,0.8168128204345703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,1,4,balanced,0.6735155487060547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,1,4,balanced,0.8784102630615234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,4,1,power_law_1.2,13.058858642578125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,1,4,balanced,0.9263526153564452
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,1,4,balanced,1.0711270141601563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,1,4,balanced,1.200769271850586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,1,4,balanced,0.02941440105438232
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,4,1,power_law_1.2,5.183048706054687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,1,4,balanced,0.03720319986343384
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,1,4,balanced,0.08533120155334473
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,1,4,balanced,0.13438207626342774
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,1,4,balanced,0.24129024505615235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,1,4,balanced,0.24028287887573244
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,1,4,balanced,0.2676851272583008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,1,4,balanced,0.267063045501709
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,1,4,balanced,0.05028992176055909
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,1,4,balanced,0.26890623092651367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,1,4,balanced,1.7128012084960935
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,1,4,balanced,0.2690048027038574
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,1,4,balanced,0.272993278503418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,1,4,balanced,0.2785740852355957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,1,4,balanced,0.2433791923522949
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,1,4,balanced,0.28224256515502927
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,1,4,balanced,0.2940876770019531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,1,4,balanced,0.32674560546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,1,4,balanced,3.1880474853515626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,1,4,balanced,0.3346227264404297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,1,4,balanced,0.3752870559692383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,1,4,balanced,0.417973747253418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,1,4,balanced,0.2697996711730957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,1,4,balanced,0.47556224822998044
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,1,4,balanced,0.5686873626708985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,1,4,power_law_1.01,0.07087359905242921
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,1,4,balanced,4.177378540039062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,1,4,balanced,0.7937088012695312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,1,4,power_law_1.01,0.12298239707946776
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,1,4,balanced,0.9307276916503906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,1,4,balanced,2.1801612854003904
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,1,4,power_law_1.01,0.19743360519409178
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,1,4,power_law_1.01,0.22360319137573242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,1,4,power_law_1.01,0.29720191955566405
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,1,4,power_law_1.01,0.2957670402526856
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,1,4,balanced,1.3646156311035156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,1,4,power_law_1.01,0.406352653503418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,1,4,power_law_1.01,0.39262847900390624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,1,4,power_law_1.01,0.5677888107299804
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,1,4,power_law_1.01,0.5848422241210938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,1,4,power_law_1.01,0.6239846420288087
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,1,4,power_law_1.01,0.6206284713745117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,1,4,power_law_1.01,0.6173811340332032
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,1,4,power_law_1.01,0.6505267333984375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,1,4,power_law_1.01,0.6748262023925781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,1,4,power_law_1.01,0.6629222106933594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,1,4,power_law_1.01,0.6937792205810547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,1,4,power_law_1.01,0.7728870391845704
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,1,4,power_law_1.01,0.8022489929199219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,1,4,power_law_1.01,0.8820019531250001
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,1,4,power_law_1.01,0.9830323028564454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,1,4,power_law_1.01,1.2794226837158202
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,1,4,power_law_1.01,0.04867839813232422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,1,4,power_law_1.01,1.4854899597167968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,1,4,power_law_1.01,0.08127103805541992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,1,4,power_law_1.01,0.09591039657592773
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,1,4,balanced,1.7658738708496096
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,1,4,power_law_1.01,0.13003647804260254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,1,4,power_law_1.01,0.1604863929748535
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,1,4,power_law_1.01,0.17064447402954103
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,1,4,power_law_1.01,0.22120191574096681
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,1,4,power_law_1.01,2.184121551513672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,1,4,power_law_1.01,0.22229375839233398
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,1,4,power_law_1.01,0.24110080718994142
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,1,4,power_law_1.01,0.24544639587402345
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,1,4,power_law_1.01,0.24639104843139648
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,1,4,power_law_1.01,0.2542617607116699
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,1,4,power_law_1.01,0.25829376220703126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,1,4,power_law_1.01,2.7126641845703126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,1,4,power_law_1.01,0.2554867172241211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,1,4,power_law_1.01,0.27487104415893554
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,1,4,power_law_1.01,0.31389440536499025
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,1,4,power_law_1.01,0.3306278228759766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,1,4,power_law_1.01,0.3874419021606445
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,1,4,power_law_1.01,0.4267993545532226
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,1,4,power_law_1.01,0.5334630584716796
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,1,4,power_law_1.01,0.2458304023742676
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,1,4,power_law_1.01,3.7149157714843755
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,1,4,power_law_1.01,0.26478975296020507
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,1,4,power_law_1.01,0.6523161315917969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,1,4,power_law_1.2,0.0708569622039795
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,1,4,power_law_1.01,0.8826815795898437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,1,4,power_law_1.2,0.11716992378234863
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,1,4,power_law_1.2,0.16490751266479492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,1,4,power_law_1.2,0.2076518440246582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,1,4,power_law_1.01,1.150382080078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,1,4,power_law_1.2,0.27183616638183594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,1,4,power_law_1.2,0.28129791259765624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,1,4,power_law_1.2,0.3818137741088867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,1,4,power_law_1.01,5.028225402832031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,1,4,power_law_1.2,0.38843006134033203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,1,4,power_law_1.01,1.5287014770507814
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,1,4,power_law_1.2,0.5758374404907227
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,1,4,power_law_1.2,0.5930879974365235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,1,4,power_law_1.2,0.6116684722900391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,1,4,power_law_1.2,0.6173465728759766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,1,4,power_law_1.2,0.6375206375122071
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,1,4,power_law_1.01,2.1083161926269534
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,1,4,power_law_1.2,0.6534950256347656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,1,4,power_law_1.2,0.6747122955322266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,1,4,power_law_1.2,0.6799001312255859
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,1,4,power_law_1.2,0.7248281860351563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,1,4,power_law_1.2,0.7750054168701171
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,1,4,power_law_1.2,0.8301862335205078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,1,4,power_law_1.2,0.9504422760009765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,1,4,power_law_1.2,1.0553638458251953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,1,4,power_law_1.2,0.04823296070098877
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,1,4,power_law_1.2,1.3908761596679686
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,1,4,power_law_1.2,0.07862016201019287
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,1,4,power_law_1.2,0.08909184455871581
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,1,4,power_law_1.2,1.5910873413085938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,1,4,power_law_1.2,0.11962623596191406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,1,4,power_law_1.2,0.1428108787536621
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,1,4,power_law_1.2,0.1749567985534668
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,1,4,power_law_1.2,0.20925823211669922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,1,4,power_law_1.2,0.2090176010131836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,1,4,power_law_1.2,2.205350341796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,1,4,power_law_1.2,0.23289344787597654
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,1,4,power_law_1.2,0.24252416610717775
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,1,4,power_law_1.2,0.2586432075500488
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,1,4,power_law_1.2,0.2596915245056152
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,1,4,power_law_1.2,0.26350080490112304
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,1,4,power_law_1.2,3.0660671997070312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,1,4,power_law_1.2,0.27853439331054686
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,1,4,power_law_1.2,0.23891328811645507
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,1,4,power_law_1.2,0.32396926879882815
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,1,4,power_law_1.2,0.2490483283996582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,1,4,power_law_1.2,0.399683837890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,1,4,power_law_1.2,0.4500902557373047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,1,4,power_law_1.2,4.118604736328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,1,4,power_law_1.2,0.5581209564208984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,1,4,power_law_1.2,0.26816640853881835
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,1,4,power_law_1.2,0.6975296020507813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,1,4,power_law_1.2,0.9369216156005858
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,1,4,power_law_1.2,0.351459846496582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,2,2,balanced,0.07304319858551026
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,1,4,power_law_1.2,1.210076141357422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,1,4,power_law_1.2,5.3714971923828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,1,4,power_law_1.2,1.699358673095703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,2,2,balanced,0.42814590454101564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,2,2,balanced,0.4339174270629883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,1,4,power_law_1.2,2.0927821350097657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,2,2,balanced,0.44060928344726563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,2,2,balanced,0.4361305618286133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,2,2,balanced,0.44064510345458985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,2,2,balanced,0.44509822845458985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,2,2,balanced,0.4745036697387695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,2,2,balanced,0.4789120101928711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,2,2,balanced,0.4421529769897461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,2,2,balanced,0.0452019214630127
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,2,2,balanced,0.03194495916366577
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,2,2,balanced,0.48140415191650393
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,2,2,balanced,0.4976012802124023
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,2,2,balanced,0.5497433471679687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,2,2,balanced,0.5778892898559571
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,2,2,balanced,0.6126169586181641
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,2,2,balanced,0.66521728515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,2,2,balanced,0.13235456466674805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,2,2,balanced,0.7402931213378906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,2,2,balanced,0.025596160888671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,2,2,balanced,1.0736128234863282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,2,2,balanced,0.033841919898986814
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,2,2,balanced,0.05422207832336425
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,2,2,balanced,1.2574272155761719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,2,2,balanced,0.138023681640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,2,2,balanced,0.2460851287841797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,2,2,balanced,0.24680448532104493
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,2,2,balanced,0.2295027160644531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,2,2,balanced,1.7404736328125001
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,2,2,balanced,0.2494374465942383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,2,2,balanced,0.2495667266845703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,2,2,balanced,0.25317247390747066
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,2,2,balanced,0.2787455940246582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,2,2,balanced,2.2153395080566405
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,2,2,balanced,0.08225791931152344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,2,2,balanced,0.4380031967163086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,2,2,balanced,0.27893375396728515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,2,2,balanced,0.28272768020629885
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,2,2,balanced,0.2857049560546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,2,2,balanced,0.2914201545715332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,2,2,balanced,0.29705984115600587
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,2,2,balanced,0.31305599212646484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,2,2,balanced,0.33279487609863284
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,2,2,balanced,0.35597694396972657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,2,2,balanced,0.3938521575927735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,2,2,balanced,0.4593523025512695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,2,2,balanced,0.2492915153503418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,2,2,balanced,0.5202009582519531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,2,2,balanced,0.7617382049560547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,2,2,power_law_1.01,0.06839935779571533
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,2,2,balanced,0.8675405120849609
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,2,2,power_law_1.01,0.09953280448913573
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,2,2,power_law_1.01,0.2165862464904785
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,2,2,balanced,1.281158447265625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,2,2,power_law_1.01,0.33999744415283206
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,2,2,power_law_1.01,0.04307839870452881
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,2,2,balanced,1.744488983154297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,2,2,power_law_1.01,0.3807436752319336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,2,2,power_law_1.01,0.37270782470703123
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,2,2,power_law_1.01,0.37139072418212893
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,2,2,power_law_1.01,0.38022655487060547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,2,2,power_law_1.01,0.39568897247314455
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,2,2,power_law_1.01,0.41323265075683596
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,2,2,power_law_1.01,0.43098495483398436
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,2,2,power_law_1.01,0.4686553573608398
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,2,2,power_law_1.01,0.46645759582519536
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,2,2,power_law_1.01,0.4854399871826172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,2,2,power_law_1.01,0.525588493347168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,2,2,power_law_1.01,0.575865592956543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,2,2,power_law_1.01,0.6460147094726563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,2,2,power_law_1.01,0.8220095825195312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,2,2,power_law_1.01,0.9397401428222656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,2,2,power_law_1.01,1.2894053649902344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,2,2,power_law_1.01,0.15777152061462402
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,2,2,power_law_1.01,0.03656064033508301
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,2,2,power_law_1.01,1.4806796264648436
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,2,2,power_law_1.01,0.05048064231872559
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,2,2,balanced,0.4688345718383789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,2,2,power_law_1.01,0.06738815784454347
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,2,2,power_law_1.01,0.10409728050231934
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,2,2,power_law_1.01,0.12845439910888673
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,2,2,power_law_1.01,0.15022336006164552
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,2,2,power_law_1.01,0.19846656799316406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,2,2,power_law_1.01,0.21196672439575198
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,2,2,power_law_1.01,0.21510271072387693
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,2,2,power_law_1.01,2.111955261230469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,2,2,power_law_1.01,0.25068799972534184
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,2,2,power_law_1.01,0.21483903884887695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,2,2,power_law_1.01,0.2163711929321289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,2,2,power_law_1.01,0.24566144943237306
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,2,2,power_law_1.01,0.2448678398132324
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,2,2,power_law_1.01,0.26681343078613284
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,2,2,power_law_1.01,0.27176191329956056
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,2,2,power_law_1.01,0.264638729095459
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,2,2,power_law_1.01,0.274968318939209
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,2,2,power_law_1.01,2.568098449707031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,2,2,power_law_1.01,0.3001241683959961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,2,2,power_law_1.01,0.3290022277832031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,2,2,power_law_1.01,0.371409912109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,2,2,power_law_1.01,0.4174796676635742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,2,2,power_law_1.01,0.5107968139648438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,2,2,power_law_1.2,0.04296703815460205
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,2,2,power_law_1.01,0.6255206298828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,2,2,power_law_1.2,0.06678400039672852
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,2,2,power_law_1.2,0.09948800086975099
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,2,2,power_law_1.01,1.0304589080810547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,2,2,power_law_1.2,0.1977894401550293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,2,2,power_law_1.2,0.236942081451416
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,2,2,power_law_1.01,1.439921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,2,2,power_law_1.2,0.3713139343261719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,2,2,power_law_1.01,0.44372608184814455
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,2,2,power_law_1.01,0.8227200317382813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,2,2,power_law_1.2,0.368908805847168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,2,2,power_law_1.2,0.14462976455688475
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,2,2,power_law_1.2,0.3962112045288086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,2,2,power_law_1.2,0.4006387329101562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,2,2,power_law_1.2,0.43604991912841795
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,2,2,power_law_1.2,0.3293388748168945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,2,2,power_law_1.2,0.48003200531005863
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,2,2,power_law_1.01,1.920305938720703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,2,2,power_law_1.2,0.47468929290771483
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,2,2,power_law_1.2,0.4838694381713867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,2,2,power_law_1.2,0.5505152130126953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,2,2,power_law_1.2,0.6081305694580078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,2,2,power_law_1.2,0.6790822601318359
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,2,2,power_law_1.2,0.8273919677734375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,2,2,power_law_1.2,0.9285568237304688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,2,2,power_law_1.2,0.036229119300842286
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,2,2,power_law_1.2,1.3169535827636718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,2,2,power_law_1.2,0.050750718116760254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,2,2,power_law_1.2,0.07173888206481933
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,2,2,power_law_1.2,1.514845428466797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,2,2,power_law_1.2,0.09262335777282715
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,2,2,power_law_1.2,0.11642623901367188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,2,2,power_law_1.2,0.14424832344055177
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,2,2,power_law_1.2,0.1914508819580078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,2,2,power_law_1.2,0.19989631652832032
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,2,2,power_law_1.2,2.204272613525391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,2,2,power_law_1.2,0.20501888275146482
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,2,2,power_law_1.2,0.21053056716918944
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,2,2,power_law_1.2,0.21192447662353517
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,2,2,power_law_1.2,0.24985855102539062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,2,2,power_law_1.2,0.25723264694213865
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,2,2,power_law_1.2,0.258288631439209
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,2,2,power_law_1.2,0.2653568077087402
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,2,2,power_law_1.2,2.681517944335938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,2,2,power_law_1.2,0.27056255340576174
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,2,2,power_law_1.2,0.2874687957763672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,2,2,power_law_1.2,0.36376190185546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,2,2,power_law_1.2,0.3104396820068359
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,2,2,power_law_1.2,0.3325811386108398
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,2,2,power_law_1.2,0.3905292892456055
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,2,2,power_law_1.2,0.43418624877929685
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,2,2,power_law_1.2,0.3911142349243164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,2,2,power_law_1.2,0.5299417495727539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,2,2,power_law_1.2,0.6458675384521484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,4,1,balanced,0.04483520030975342
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,4,1,balanced,0.057377920150756836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,2,2,power_law_1.2,0.8428646087646484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,4,1,balanced,0.07936895847320556
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,4,1,balanced,0.13288960456848145
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,2,2,power_law_1.2,1.0428607940673829
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,4,1,balanced,0.2370207977294922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,4,1,balanced,0.44396095275878905
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,4,1,balanced,0.44723072052001955
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,2,2,power_law_1.2,1.4776332092285156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,4,1,balanced,0.4497548675537109
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,4,1,balanced,0.4526150512695312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,2,2,power_law_1.2,1.875751647949219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,4,1,balanced,0.4543884658813477
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,4,1,balanced,0.457823371887207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,4,1,balanced,0.46415489196777343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,4,1,balanced,0.46573249816894535
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,4,1,balanced,0.47376575469970705
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,4,1,balanced,0.6643443298339844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,4,1,balanced,0.6696134185791015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,4,1,balanced,0.682265625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,4,1,balanced,0.7202470397949219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,4,1,balanced,0.7517017364501953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,4,1,balanced,0.8134060668945313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,4,1,balanced,0.8680703735351563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,4,1,balanced,0.03209919929504394
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,4,1,balanced,1.0495500946044922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,4,1,balanced,1.1568473815917968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,4,1,balanced,0.09746879577636719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,4,1,balanced,1.593868865966797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,4,1,balanced,0.15615615844726563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,4,1,balanced,0.04501567840576172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,4,1,balanced,0.2772550392150879
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,4,1,balanced,0.06223296165466309
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,4,1,balanced,0.2793215942382813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,4,1,balanced,0.27886592864990234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,4,1,balanced,0.2814143943786621
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,2,2,power_law_1.2,0.4815552139282227
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,4,1,balanced,0.2833625602722168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,4,1,balanced,0.2851788711547852
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,4,1,balanced,0.2867622375488281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,4,1,balanced,0.28983488082885744
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,4,1,balanced,0.2934796714782715
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,4,1,balanced,0.3359833526611328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,4,1,balanced,0.3385580825805664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,4,1,balanced,2.0586119079589844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,4,1,balanced,0.3415929412841797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,4,1,balanced,0.3538739013671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,4,1,balanced,2.9645068359374998
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,4,1,balanced,0.36044544219970703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,4,1,balanced,0.3864083099365234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,4,1,balanced,0.41085632324218746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,4,1,balanced,0.46724609375000004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,4,1,balanced,0.5216614532470703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,4,1,power_law_1.01,0.04630784034729004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,4,1,balanced,0.7976825714111329
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,4,1,power_law_1.01,0.05657343864440918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,4,1,balanced,3.832469787597656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,4,1,balanced,0.9038547515869141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,4,1,power_law_1.01,0.07972608089447022
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,4,1,power_law_1.01,0.13295424461364747
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,4,1,power_law_1.01,0.20251327514648437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,4,1,power_law_1.01,0.2401817512512207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,4,1,power_law_1.01,0.33612865447998047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,4,1,power_law_1.01,0.3500422286987305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,4,1,power_law_1.01,0.3577888107299805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,4,1,power_law_1.01,0.36512447357177735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,4,1,power_law_1.01,0.3951007843017578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,4,1,power_law_1.01,0.40712512969970704
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,4,1,power_law_1.01,0.41146686553955075
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,4,1,power_law_1.01,0.44181697845458984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,4,1,power_law_1.01,0.5926188659667969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,4,1,power_law_1.01,0.600719985961914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,4,1,balanced,1.3191065979003906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,4,1,power_law_1.01,0.6255046463012695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,4,1,power_law_1.01,0.6721119689941406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,4,1,balanced,1.6979603576660156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,4,1,power_law_1.01,0.7161190032958984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,4,1,power_law_1.01,0.7873683166503906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,4,1,power_law_1.01,0.8565740966796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,4,1,power_law_1.01,0.03218816041946411
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,4,1,power_law_1.01,1.0336089324951172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,4,1,power_law_1.01,0.04668735980987549
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,4,1,power_law_1.01,0.06070464134216309
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,4,1,power_law_1.01,1.2408358764648437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,4,1,power_law_1.01,0.09704319953918457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,4,1,power_law_1.01,0.13180607795715332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,4,1,power_law_1.01,0.21707712173461916
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,4,1,power_law_1.01,1.6855328369140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,4,1,power_law_1.01,0.2232320022583008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,4,1,power_law_1.01,0.22759103775024414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,4,1,power_law_1.01,0.23203775405883792
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,4,1,power_law_1.01,2.1559698486328123
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,4,1,power_law_1.01,0.1586124801635742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,4,1,power_law_1.01,0.24817407608032224
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,4,1,power_law_1.01,0.2511027145385742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,4,1,power_law_1.01,0.2612607955932617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,4,1,power_law_1.01,0.30560319900512695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,4,1,power_law_1.01,0.30521215438842775
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,4,1,power_law_1.01,0.24388671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,4,1,power_law_1.01,0.3221484756469727
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,4,1,power_law_1.01,0.340780143737793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,4,1,power_law_1.01,3.024852600097656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,4,1,power_law_1.01,0.3555807876586914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,4,1,power_law_1.01,0.4009280014038086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,4,1,power_law_1.01,0.4445926284790039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,4,1,power_law_1.01,0.5356620788574219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,4,1,power_law_1.01,0.6335686492919922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,4,1,power_law_1.2,0.045632638931274415
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,4,1,power_law_1.01,4.008084411621094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,4,1,power_law_1.01,0.8265650939941406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,4,1,power_law_1.2,0.057018880844116215
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,4,1,power_law_1.2,0.07979072093963623
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,4,1,power_law_1.2,0.12375103950500488
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,4,1,power_law_1.01,1.0203257751464843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,4,1,power_law_1.2,0.2009894371032715
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,4,1,power_law_1.2,0.22987136840820313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,4,1,power_law_1.2,0.3293907165527344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,4,1,power_law_1.2,0.3438681411743164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,4,1,power_law_1.2,0.3516377639770508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,4,1,power_law_1.01,1.4330355834960937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,4,1,power_law_1.2,0.370552978515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,4,1,power_law_1.2,0.3929286575317383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,4,1,power_law_1.2,0.40288127899169923
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,4,1,power_law_1.2,0.4119404983520508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,4,1,power_law_1.2,0.450450553894043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,4,1,power_law_1.01,1.8531712341308595
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,4,1,power_law_1.2,0.5957632064819336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,4,1,power_law_1.2,0.6127622222900391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,4,1,power_law_1.2,0.6354336166381837
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,4,1,power_law_1.2,0.7059014129638672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,4,1,power_law_1.2,0.729784927368164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,4,1,power_law_1.2,0.8158854675292968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,4,1,power_law_1.2,0.885555191040039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,4,1,power_law_1.2,0.033148798942565914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,4,1,power_law_1.2,1.0745247650146486
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,4,1,power_law_1.2,0.04712768077850342
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,4,1,power_law_1.2,0.061977601051330565
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,4,1,power_law_1.2,1.2942189025878907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,4,1,power_law_1.2,0.09210047721862794
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,4,1,power_law_1.2,0.12095295906066894
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,4,1,power_law_1.2,0.15073599815368652
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,4,1,power_law_1.2,0.22113088607788084
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,4,1,power_law_1.2,0.21413503646850587
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,4,1,power_law_1.2,0.23084543228149412
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,4,1,power_law_1.2,0.22403263092041015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,4,1,power_law_1.2,2.1700396728515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,4,1,power_law_1.2,0.23785919189453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,4,1,power_law_1.2,0.24519807815551758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,4,1,power_law_1.2,0.2514899253845215
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,4,1,power_law_1.2,0.2608518409729004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,4,1,power_law_1.2,0.3194131278991699
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,4,1,power_law_1.2,1.7312031555175782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,4,1,power_law_1.2,0.32426368713378906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,4,1,power_law_1.2,0.34982208251953123
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,4,1,power_law_1.2,0.3746547317504883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,4,1,power_law_1.2,3.0980609130859373
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,4,1,power_law_1.2,0.45977535247802737
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,4,1,power_law_1.2,0.5577107238769531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,4,1,power_law_1.2,0.3257612609863281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,1,4,balanced,0.04358399868011474
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,4,1,power_law_1.2,0.6622617340087891
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,1,4,balanced,0.04899456024169922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,1,4,balanced,0.11050623893737792
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,4,1,power_law_1.2,3.9752069091796876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,4,1,power_law_1.2,0.839375991821289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,4,1,power_law_1.2,0.4175904083251953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,1,4,balanced,0.2069811248779297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,1,4,balanced,0.26595071792602537
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,4,1,power_law_1.2,1.074988784790039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,1,4,balanced,0.3762035369873047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,1,4,balanced,0.37468929290771485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,1,4,balanced,0.3802560043334961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,1,4,balanced,0.3843494415283203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,1,4,balanced,0.3966540908813477
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,1,4,balanced,0.39960575103759766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,4,1,power_law_1.2,1.4617535400390627
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,1,4,balanced,0.3986713409423828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,1,4,balanced,0.39989887237548827
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,1,4,balanced,0.41385345458984374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,1,4,balanced,0.4278054428100586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,1,4,balanced,0.46038017272949217
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,1,4,balanced,0.06414847850799561
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,4,1,power_law_1.2,1.9053919982910155
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,1,4,balanced,0.5308556747436524
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,1,4,balanced,0.5753676986694336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,1,4,balanced,0.39149696350097657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,1,4,balanced,0.7996940612792969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,1,4,power_law_1.01,0.06275968074798584
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,1,4,balanced,0.910640640258789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,1,4,power_law_1.01,0.09389439582824707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,1,4,power_law_1.01,0.10943103790283204
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,1,4,balanced,0.4800870513916015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,1,4,power_law_1.01,0.1796940803527832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,1,4,balanced,1.3463629150390626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,1,4,power_law_1.01,0.334095344543457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,1,4,power_law_1.01,0.33748607635498046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,1,4,power_law_1.01,0.3378688049316406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,1,4,power_law_1.01,0.3453657531738281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,1,4,balanced,2.4488729858398437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,1,4,power_law_1.01,0.3597248077392578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,1,4,power_law_1.01,0.3738111877441406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,1,4,balanced,1.6765081787109373
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,1,4,power_law_1.01,0.3826841735839844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,1,4,power_law_1.01,0.3942451095581055
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,1,4,power_law_1.01,0.40313983917236323
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,1,4,balanced,3.292798767089844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,1,4,power_law_1.01,0.4414630508422851
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,1,4,power_law_1.01,0.4504819107055664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,1,4,power_law_1.01,0.5056063842773437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,1,4,power_law_1.01,0.5688742446899414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,1,4,power_law_1.01,0.6854438018798829
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,1,4,power_law_1.01,0.9299890899658203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,1,4,power_law_1.01,0.17480960845947266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,1,4,power_law_1.01,0.237127685546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,1,4,power_law_1.01,1.1559833526611327
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,1,4,power_law_1.2,0.06267903804779053
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,1,4,power_law_1.2,0.07756159782409669
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,1,4,power_law_1.2,0.11870464324951172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,1,4,power_law_1.2,0.1647897529602051
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,1,4,power_law_1.2,0.18338560104370116
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,1,4,power_law_1.01,1.6833638000488282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,1,4,power_law_1.2,0.3367308807373047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,1,4,power_law_1.2,0.34717823028564454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,1,4,power_law_1.01,2.179352264404297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,1,4,power_law_1.2,0.3510976028442383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,1,4,power_law_1.2,0.37942783355712895
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,1,4,power_law_1.2,0.38833023071289063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,1,4,power_law_1.2,0.3865331268310547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,1,4,power_law_1.01,2.9930932617187502
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,1,4,power_law_1.01,0.41402240753173825
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,1,4,power_law_1.2,0.4144704055786133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,1,4,power_law_1.2,0.4146342468261719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,1,4,power_law_1.2,0.42576511383056637
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,1,4,power_law_1.2,0.4499878311157227
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,1,4,power_law_1.2,0.5143948745727539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,1,4,power_law_1.2,0.6012684631347656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,1,4,power_law_1.2,0.39430015563964843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,1,4,power_law_1.2,0.7112268829345704
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,1,4,power_law_1.01,4.253207092285156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,1,4,power_law_1.2,1.0009407806396484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,1,4,power_law_1.2,0.23776512145996093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,2,2,balanced,0.043888640403747556
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,2,2,balanced,0.06214272022247315
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,1,4,power_law_1.2,1.3533990478515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,2,2,balanced,0.10911871910095214
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,1,4,power_law_1.2,0.3370240020751953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,2,2,balanced,0.1897420883178711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,2,2,balanced,0.260318717956543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,2,2,balanced,0.2655155181884766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,2,2,balanced,0.2675904083251953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,2,2,balanced,0.3563251113891602
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,2,2,balanced,0.35849857330322266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,2,2,balanced,0.36526462554931644
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,2,2,balanced,0.36821502685546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,2,2,balanced,0.034766080379486086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,2,2,balanced,0.37275646209716795
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,2,2,balanced,0.3850777435302734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,1,4,power_law_1.2,2.405193023681641
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,2,2,balanced,0.38548095703124996
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,2,2,balanced,0.39756031036376954
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,2,2,balanced,0.40941310882568355
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,2,2,balanced,0.4353177642822265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,2,2,balanced,0.451701774597168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,2,2,balanced,0.516189422607422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,2,2,balanced,0.5599462509155273
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,2,2,balanced,0.7628070068359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,2,2,balanced,0.8688639831542968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,1,4,power_law_1.2,1.8760704040527343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,2,2,power_law_1.01,0.04382080078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,2,2,power_law_1.01,0.07325183868408203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,2,2,balanced,1.281871337890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,2,2,power_law_1.01,0.12162176132202149
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,2,2,power_law_1.01,0.14932479858398437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,1,4,power_law_1.2,3.524666748046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,2,2,power_law_1.01,0.208090877532959
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,2,2,power_law_1.01,0.22621952056884767
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,2,2,power_law_1.01,0.3246361541748047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,2,2,balanced,2.310055694580078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,2,2,power_law_1.01,0.3246246337890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,2,2,power_law_1.01,0.3369075012207031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,2,2,balanced,1.5982028198242186
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,2,2,power_law_1.01,0.3466815948486328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,2,2,power_law_1.01,0.35976703643798824
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,2,2,balanced,2.9955316162109376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,2,2,power_law_1.01,0.3717619323730469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,2,2,power_law_1.01,0.3777548980712891
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,2,2,power_law_1.01,0.3894668960571289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,2,2,power_law_1.01,0.41520511627197265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,2,2,power_law_1.01,0.43773056030273433
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,2,2,power_law_1.01,0.45142654418945316
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,2,2,power_law_1.01,0.5297151947021485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,2,2,power_law_1.01,0.22543359756469727
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,2,2,power_law_1.01,0.05769983768463135
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,2,2,power_law_1.01,0.6157977676391602
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,2,2,power_law_1.01,0.7973990631103516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,2,2,power_law_1.01,1.0207513427734374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,2,2,power_law_1.2,0.052930560111999515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,2,2,power_law_1.01,1.2987545776367189
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,2,2,power_law_1.2,0.0747046422958374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,2,2,power_law_1.2,0.10572287559509277
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,2,2,power_law_1.01,1.8012658691406251
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,2,2,power_law_1.2,0.13653120040893554
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,2,2,power_law_1.2,0.20759935379028321
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,2,2,power_law_1.2,0.21267583847045896
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,2,2,power_law_1.2,0.04325119972229004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,2,2,power_law_1.2,0.3221350479125976
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,2,2,power_law_1.01,2.5072230529785156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,2,2,power_law_1.2,0.33060352325439457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,1,4,power_law_1.2,5.068392944335938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,2,2,power_law_1.2,0.34863616943359377
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,2,2,power_law_1.2,0.35454334259033204
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,2,2,power_law_1.2,0.362979850769043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,2,2,power_law_1.2,0.3776601409912109
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,2,2,power_law_1.2,0.3922111892700195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,2,2,power_law_1.2,0.40091777801513667
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,2,2,power_law_1.2,0.4239961624145508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,2,2,power_law_1.01,3.573197937011719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,2,2,power_law_1.2,0.4708070373535156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,2,2,power_law_1.2,0.5368064117431641
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,2,2,power_law_1.2,0.6379084777832031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,2,2,power_law_1.2,0.23375232696533202
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,2,2,power_law_1.2,0.8331699371337891
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,4,1,balanced,0.03624576091766357
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,4,1,balanced,0.05239488124847412
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,2,2,power_law_1.2,1.0668160247802736
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,4,1,balanced,0.0721068811416626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,2,2,power_law_1.2,0.4387481689453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,4,1,balanced,0.11582847595214844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,4,1,balanced,0.20078079223632814
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,2,2,power_law_1.2,1.3817164611816406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,4,1,balanced,0.2720479965209961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,4,1,balanced,0.27501951217651366
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,4,1,balanced,0.2766547203063965
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,4,1,balanced,0.27795455932617186
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,4,1,balanced,0.2811699295043945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,4,1,balanced,0.2816230392456055
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,2,2,power_law_1.2,1.8622950744628906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,4,1,balanced,0.3742156982421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,4,1,balanced,0.3805120086669922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,4,1,balanced,0.3880518341064453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,2,2,power_law_1.2,2.484467163085937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,4,1,balanced,0.3901145553588867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,4,1,balanced,0.3991667175292969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,4,1,balanced,0.4340243148803711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,4,1,balanced,0.44562686920166017
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,4,1,balanced,0.41098751068115236
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,4,1,balanced,0.5434048080444336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,4,1,balanced,0.7899078369140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,4,1,balanced,0.8983103942871093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,4,1,power_law_1.01,0.03676032066345215
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,4,1,power_law_1.01,0.05122623920440674
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,4,1,power_law_1.01,0.07297728061676026
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,4,1,balanced,1.3427212524414063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,4,1,power_law_1.01,0.10302080154418944
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,4,1,power_law_1.01,0.15331392288208007
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,4,1,balanced,1.6720941162109377
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,4,1,power_law_1.01,0.20538560867309572
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,4,1,power_law_1.01,0.2172934341430664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,4,1,power_law_1.01,0.22849664688110352
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,4,1,balanced,0.5829420852661132
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,4,1,power_law_1.01,0.24461055755615235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,4,1,power_law_1.01,0.2560531234741211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,4,1,power_law_1.01,0.2677996826171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,4,1,power_law_1.01,0.34751678466796876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,4,1,balanced,2.432106170654297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,4,1,power_law_1.01,0.35750144958496094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,4,1,power_law_1.01,0.3709747314453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,4,1,power_law_1.01,0.37794559478759765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,4,1,power_law_1.01,0.3862374496459961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,4,1,power_law_1.01,0.40510337829589843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,4,1,power_law_1.01,0.43342334747314454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,4,1,power_law_1.01,0.45203712463378903
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,4,1,power_law_1.01,0.5157555389404297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,2,2,power_law_1.2,3.7888473510742187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,4,1,power_law_1.01,0.5927564620971679
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,4,1,power_law_1.01,0.7648441314697265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,4,1,power_law_1.2,0.03591871976852417
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,4,1,power_law_1.2,0.05069568157196045
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,4,1,power_law_1.2,0.07190400123596191
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,4,1,power_law_1.01,0.9528057861328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,4,1,power_law_1.2,0.0963987159729004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,4,1,balanced,3.184228515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,4,1,power_law_1.2,0.12944191932678223
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,4,1,power_law_1.2,0.20518144607543945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,4,1,power_law_1.01,1.3298098754882812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,4,1,power_law_1.2,0.218920955657959
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,4,1,power_law_1.2,0.23658815383911133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,4,1,power_law_1.2,0.2510508728027344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,4,1,power_law_1.2,0.26211584091186524
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,4,1,power_law_1.01,1.702309112548828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,4,1,power_law_1.2,0.3575993728637695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,4,1,power_law_1.2,0.3648691177368164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,4,1,power_law_1.2,0.3803104019165039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,4,1,power_law_1.2,0.3883859252929688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,4,1,power_law_1.2,0.39918529510498046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,4,1,power_law_1.2,0.4208671951293946
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,4,1,power_law_1.2,0.4389068984985352
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,4,1,power_law_1.2,0.4556281661987304
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,4,1,power_law_1.2,0.5245811080932616
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,4,1,power_law_1.2,0.2684518432617188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,4,1,power_law_1.2,0.6046931076049804
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,4,1,power_law_1.2,0.7798035430908203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,4,1,power_law_1.01,2.467008056640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,1,4,balanced,0.11657216072082519
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,4,1,power_law_1.2,0.9613843536376953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,1,4,balanced,0.23872512817382813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,1,4,balanced,0.40913150787353514
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,1,4,balanced,0.7501222229003905
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,4,1,power_law_1.01,3.224435119628906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,4,1,power_law_1.2,1.3375430297851563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,1,4,balanced,1.4251072692871094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,1,4,balanced,1.9768307495117188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,4,1,power_law_1.2,1.7201586914062499
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,1,4,balanced,1.9972108459472657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,1,4,balanced,2.0049650573730466
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,1,4,balanced,2.0274394226074217
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,1,4,balanced,2.045244140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,1,4,balanced,2.0552359008789063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,1,4,balanced,3.1062744140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,4,1,power_law_1.2,2.484962615966797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,1,4,balanced,3.1476788330078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,1,4,balanced,3.218362731933594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,1,4,balanced,3.2416613769531253
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,1,4,balanced,3.318289794921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,1,4,balanced,3.420769348144531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,1,4,balanced,3.669253234863281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,1,4,balanced,3.846842956542969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,1,4,balanced,4.216482543945313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,4,1,power_law_1.2,3.230037841796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,1,4,balanced,4.511659545898437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,1,4,balanced,0.07788544178009034
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,1,4,balanced,0.1377676773071289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,1,4,balanced,0.23738111495971678
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,1,4,balanced,0.3987609481811524
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,1,4,balanced,0.7219302368164062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,1,4,balanced,6.836123046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,1,4,balanced,1.0089766693115234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,1,4,balanced,1.012490234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,1,4,balanced,1.023019485473633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,1,4,balanced,7.957615356445312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,1,4,balanced,1.0362419128417968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,1,4,balanced,1.0435340881347657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,1,4,balanced,1.0625049591064453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,1,4,balanced,1.1246733093261718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,1,4,balanced,1.1333721923828126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,1,4,balanced,1.1975027465820314
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,1,4,balanced,1.1850943756103516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,1,4,balanced,1.211141128540039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,1,4,balanced,1.2605363464355468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,1,4,balanced,1.3272000122070313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,1,4,balanced,1.412604217529297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,1,4,balanced,1.5660032653808593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,1,4,balanced,12.460505371093749
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,1,4,balanced,1.718233642578125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,1,4,balanced,2.4930496215820312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,1,4,balanced,2.8445120239257813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,1,4,power_law_1.01,0.38427265167236324
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,1,4,balanced,4.309698486328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,1,4,balanced,15.524549560546877
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,1,4,power_law_1.01,0.7056243133544922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,1,4,balanced,4.721479797363282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,1,4,power_law_1.01,0.7581017303466797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,1,4,power_law_1.01,0.9693401336669922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,1,4,power_law_1.01,1.2071475219726562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,1,4,power_law_1.01,1.7239680480957031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,1,4,power_law_1.01,1.6121472167968751
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,1,4,power_law_1.01,1.6617741394042969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,1,4,balanced,7.431489257812499
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,1,4,power_law_1.01,1.6992819213867185
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,1,4,power_law_1.01,1.7580070495605469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,1,4,power_law_1.01,2.768663024902344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,1,4,power_law_1.01,2.8550399780273437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,1,4,balanced,9.085331420898438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,1,4,power_law_1.01,2.9410086059570313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,1,4,power_law_1.01,3.0786648559570313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,1,4,balanced,23.675556640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,1,4,power_law_1.01,1.875855407714844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,1,4,power_law_1.01,3.200899963378906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,1,4,power_law_1.01,3.3705560302734376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,1,4,power_law_1.01,3.6998593139648435
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,1,4,power_law_1.01,4.019319152832031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,1,4,power_law_1.01,5.251902465820312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,1,4,balanced,29.9934765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,1,4,power_law_1.01,6.993117675781249
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,1,4,power_law_1.01,0.21918848037719724
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,1,4,power_law_1.01,0.388540153503418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,1,4,power_law_1.01,0.39497726440429687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,1,4,power_law_1.01,0.49678207397460933
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,1,4,power_law_1.01,4.656422424316406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,1,4,power_law_1.01,8.404963989257812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,1,4,power_law_1.01,0.5796838378906249
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,1,4,power_law_1.01,0.8884313964843751
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,1,4,power_law_1.01,0.8247052764892577
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,1,4,power_law_1.01,0.8640447998046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,1,4,power_law_1.01,0.8970188903808595
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,1,4,power_law_1.01,0.9237696075439453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,1,4,power_law_1.01,0.9394739532470704
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,1,4,power_law_1.01,1.0144306945800783
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,1,4,power_law_1.01,1.0259942626953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,1,4,power_law_1.01,1.1050342559814452
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,1,4,power_law_1.01,12.5197509765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,1,4,power_law_1.01,1.0996543884277343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,1,4,power_law_1.01,1.1198540496826173
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,1,4,power_law_1.01,1.1901913452148438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,1,4,power_law_1.01,1.319066925048828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,1,4,power_law_1.01,1.4362803649902343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,1,4,power_law_1.01,1.718303985595703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,1,4,power_law_1.01,15.0710107421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,1,4,power_law_1.01,1.9404582214355468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,1,4,power_law_1.01,2.504596405029297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,1,4,power_law_1.2,0.3815167999267578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,1,4,power_law_1.01,3.2015066528320313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,1,4,power_law_1.2,0.5467635345458984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,1,4,power_law_1.2,0.7434585571289063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,1,4,power_law_1.01,4.260505676269531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,1,4,power_law_1.2,0.955447006225586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,1,4,power_law_1.2,1.2272576141357423
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,1,4,power_law_1.2,1.7133938598632814
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,1,4,power_law_1.01,5.52501220703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,1,4,power_law_1.2,1.6224691772460935
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,1,4,power_law_1.2,1.6533311462402345
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,1,4,power_law_1.01,22.50840087890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,1,4,power_law_1.2,1.7899481201171874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,1,4,power_law_1.01,7.938588256835938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,1,4,power_law_1.2,1.9086271667480468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,1,4,power_law_1.2,2.8400128173828127
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,1,4,power_law_1.2,2.9645785522460937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,1,4,power_law_1.01,10.465313720703126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,1,4,power_law_1.2,1.687132110595703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,1,4,power_law_1.2,3.1620391845703124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,1,4,power_law_1.2,3.18065673828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,1,4,power_law_1.2,3.307403564453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,1,4,power_law_1.2,3.5092672729492187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,1,4,power_law_1.2,4.1292416381835935
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,1,4,power_law_1.2,4.940623474121094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,1,4,power_law_1.2,5.475264282226563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,1,4,power_law_1.2,3.7845889282226564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,1,4,power_law_1.01,31.6157763671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,1,4,power_law_1.2,7.546677856445312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,1,4,power_law_1.2,0.21906303405761718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,1,4,power_law_1.2,0.32081409454345705
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,1,4,power_law_1.2,0.4180326461791992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,1,4,power_law_1.2,0.4804748916625977
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,1,4,power_law_1.2,9.152578735351563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,1,4,power_law_1.2,0.6032473754882812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,1,4,power_law_1.2,0.8752063751220703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,1,4,power_law_1.2,0.8614361572265625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,1,4,power_law_1.2,0.8487769317626952
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,1,4,power_law_1.2,0.9003289794921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,1,4,power_law_1.2,0.8987648010253906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,1,4,power_law_1.2,0.9374886322021485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,1,4,power_law_1.2,1.0366719818115233
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,1,4,power_law_1.2,1.06761474609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,1,4,power_law_1.2,1.1203123474121095
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,1,4,power_law_1.2,1.1381298828125002
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,1,4,power_law_1.2,1.155577621459961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,1,4,power_law_1.2,1.2693695831298828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,1,4,power_law_1.2,1.3433612060546873
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,1,4,power_law_1.2,1.5227699279785156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,1,4,power_law_1.2,14.085662841796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,1,4,power_law_1.2,1.7433241271972655
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,1,4,power_law_1.2,2.1061541748046873
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,1,4,power_law_1.2,2.646178588867188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,1,4,power_law_1.2,3.29765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,2,2,balanced,0.1202291202545166
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,2,2,balanced,0.23767808914184568
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,2,2,balanced,0.38233600616455077
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,1,4,power_law_1.2,17.150156250000002
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,1,4,power_law_1.2,4.958222045898437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,2,2,balanced,0.7144115447998047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,1,4,power_law_1.2,5.4735913085937495
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,2,2,balanced,1.3504856872558595
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,2,2,balanced,2.0011724853515624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,2,2,balanced,1.9834815979003906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,2,2,balanced,2.010950469970703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,2,2,balanced,2.0326527404785155
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,1,4,power_law_1.2,8.182576904296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,2,2,balanced,2.052185668945312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,2,2,balanced,2.0502105712890626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,2,2,balanced,2.095370178222656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,2,2,balanced,2.1037440490722656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,2,2,balanced,2.1323251342773437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,2,2,balanced,3.13842041015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,2,2,balanced,3.2173925781249997
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,1,4,power_law_1.2,10.734747314453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,2,2,balanced,3.336293029785156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,2,2,balanced,3.597754821777344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,1,4,power_law_1.2,26.7241015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,2,2,balanced,3.7730535888671874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,2,2,balanced,4.2587646484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,2,2,balanced,4.599967956542969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,2,2,balanced,0.0766323184967041
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,2,2,balanced,0.1306175994873047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,2,2,balanced,6.145626831054687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,2,2,balanced,6.981422119140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,2,2,balanced,0.7080140686035156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,2,2,balanced,0.22716928482055665
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,2,2,balanced,1.0178189086914062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,2,2,balanced,1.021539840698242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,2,2,balanced,1.033570556640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,2,2,balanced,1.0416575622558593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,2,2,balanced,11.102161865234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,2,2,balanced,0.39858943939208985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,1,4,power_law_1.2,35.54617431640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,2,2,balanced,1.050967025756836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,2,2,balanced,1.0648857879638673
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,2,2,balanced,1.0834598541259766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,2,2,balanced,1.0930111694335938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,2,2,balanced,1.1287449645996095
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,2,2,balanced,1.2309145355224609
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,2,2,balanced,1.2804595947265625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,2,2,balanced,1.2790386962890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,2,2,balanced,1.3723008728027344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,2,2,balanced,1.446366729736328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,2,2,balanced,13.894003906250001
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,2,2,balanced,1.5603302001953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,2,2,balanced,1.6374131774902345
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,2,2,balanced,2.571813049316406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,2,2,balanced,2.891614685058594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,2,2,power_law_1.01,0.23655424118041993
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,2,2,power_law_1.01,0.38208641052246095
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,2,2,power_law_1.01,0.47093505859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,2,2,balanced,4.271311340332032
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,2,2,power_law_1.01,0.6997337341308594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,2,2,balanced,4.822248840332032
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,2,2,power_law_1.01,0.9242752075195313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,2,2,power_law_1.01,1.5728410339355468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,2,2,power_law_1.01,1.55728515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,2,2,power_law_1.01,1.6621107482910156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,2,2,power_law_1.01,1.6528536987304687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,2,2,balanced,7.506165771484374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,2,2,power_law_1.01,1.7377561950683593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,2,2,power_law_1.01,1.8149938964843748
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,2,2,power_law_1.01,1.8894003295898436
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,2,2,power_law_1.01,2.0185229492187498
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,2,2,balanced,21.56840087890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,2,2,balanced,9.148102416992188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,2,2,power_law_1.01,2.336806335449219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,2,2,power_law_1.01,2.9917080688476565
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,2,2,power_law_1.01,3.10463623046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,2,2,power_law_1.01,3.2779531860351563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,2,2,power_law_1.01,3.4805966186523434
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,2,2,power_law_1.01,3.825430908203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,2,2,power_law_1.01,4.252368774414062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,2,2,power_law_1.01,4.846554870605469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,2,2,balanced,27.684433593749997
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,2,2,power_law_1.01,0.12636287689208986
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,2,2,power_law_1.01,0.29290624618530275
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,2,2,power_law_1.01,6.42205810546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,2,2,power_law_1.01,0.3974425506591797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,2,2,power_law_1.01,0.49272705078124995
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,2,2,power_law_1.01,0.8095513916015624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,2,2,power_law_1.01,7.615087280273437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,2,2,power_law_1.01,0.8509069061279296
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,2,2,power_law_1.01,0.8552371215820311
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,2,2,power_law_1.01,0.8703897857666016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,2,2,power_law_1.01,0.22187263488769532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,2,2,power_law_1.01,0.9075942230224608
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,2,2,power_law_1.01,0.9216204833984374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,2,2,power_law_1.01,0.9668812561035157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,2,2,power_law_1.01,1.03897216796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,2,2,power_law_1.01,10.861444091796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,2,2,power_law_1.01,1.1635916900634764
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,2,2,power_law_1.01,1.1938098907470702
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,2,2,power_law_1.01,1.2455487823486329
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,2,2,power_law_1.01,1.3243968200683593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,2,2,power_law_1.01,1.4350770568847655
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,2,2,power_law_1.01,1.643888702392578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,2,2,power_law_1.01,0.9766566467285156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,2,2,power_law_1.01,13.546033935546877
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,2,2,power_law_1.01,2.0078207397460934
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,2,2,power_law_1.01,2.4383142089843752
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,2,2,power_law_1.2,0.23918975830078124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,2,2,power_law_1.01,2.974243774414062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,2,2,power_law_1.2,0.29402240753173825
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,2,2,power_law_1.2,0.4992627334594727
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,2,2,power_law_1.01,4.005672912597656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,2,2,power_law_1.2,0.8619750213623046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,2,2,power_law_1.2,0.6485517120361328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,2,2,power_law_1.2,1.5496012878417968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,2,2,power_law_1.01,19.725346679687497
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,2,2,power_law_1.2,1.5351551818847657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,2,2,power_law_1.2,1.6149874877929686
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,2,2,power_law_1.01,7.487340698242187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,2,2,power_law_1.2,1.710743103027344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,2,2,power_law_1.2,1.7594969177246091
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,2,2,power_law_1.2,1.8582220458984373
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,2,2,power_law_1.2,1.9755891418457032
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,2,2,power_law_1.2,2.1123175048828124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,2,2,power_law_1.2,2.4361651611328123
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,2,2,power_law_1.01,9.958829956054688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,2,2,power_law_1.01,5.155034790039062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,2,2,power_law_1.2,3.036052551269531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,2,2,power_law_1.2,3.1690484619140626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,2,2,power_law_1.2,3.3237991333007812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,2,2,power_law_1.01,25.819765625000002
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,2,2,power_law_1.2,3.522128601074219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,2,2,power_law_1.2,3.825119934082031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,2,2,power_law_1.2,4.381560363769531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,2,2,power_law_1.2,0.12492544174194335
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,2,2,power_law_1.2,5.1655810546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,2,2,power_law_1.2,0.1946905517578125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,2,2,power_law_1.2,0.2861311912536621
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,2,2,power_law_1.2,6.568203735351562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,2,2,power_law_1.2,0.4854707336425781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,2,2,power_law_1.2,0.7396697235107422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,2,2,power_law_1.2,0.8431795501708985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,2,2,power_law_1.2,7.999313964843751
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,2,2,power_law_1.2,0.8562713623046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,2,2,power_law_1.2,0.8859085083007813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,2,2,power_law_1.2,0.9050994873046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,2,2,power_law_1.2,0.36372222900390627
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,2,2,power_law_1.2,0.9412108612060546
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,2,2,power_law_1.2,0.9927718353271484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,2,2,power_law_1.2,1.0174028778076172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,2,2,power_law_1.2,1.0770368194580078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,2,2,power_law_1.2,1.1644735717773438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,2,2,power_law_1.2,1.2326937866210936
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,2,2,power_law_1.2,11.715523681640624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,2,2,power_law_1.2,1.265506591796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,2,2,power_law_1.2,1.3664997863769532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,2,2,power_law_1.2,1.470031433105469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,2,2,power_law_1.2,1.7766502380371094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,2,2,power_law_1.2,2.140536346435547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,2,2,power_law_1.2,14.287413330078124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,2,2,power_law_1.2,2.5848678588867187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,2,2,power_law_1.2,3.2082342529296874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,4,1,balanced,0.12980416297912598
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,2,2,power_law_1.2,4.2045965576171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,4,1,balanced,0.4000236892700196
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,4,1,balanced,0.255599365234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,2,2,power_law_1.2,5.578367919921876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,4,1,balanced,0.732537612915039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,2,2,power_law_1.2,21.26719970703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,4,1,balanced,1.3659468078613282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,2,2,power_law_1.2,7.377539672851563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,4,1,balanced,2.035210266113281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,4,1,balanced,2.0431610107421876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,4,1,balanced,2.0415795898437503
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,2,2,power_law_1.2,9.585624389648437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,4,1,balanced,2.138165740966797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,4,1,balanced,2.0569497680664064
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,4,1,balanced,2.0676165771484376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,4,1,balanced,2.1673536682128907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,4,1,balanced,2.100189514160156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,4,1,balanced,2.1817689514160157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,2,2,power_law_1.2,27.180576171874996
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,4,1,balanced,2.7075839233398438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,4,1,balanced,3.3491717529296876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,4,1,balanced,2.209998779296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,4,1,balanced,3.5093069458007813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,4,1,balanced,3.6916717529296874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,4,1,balanced,4.207645568847656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,4,1,balanced,0.09646400451660156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,4,1,balanced,0.14757311820983887
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,4,1,balanced,4.742584838867187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,4,1,balanced,0.2365235137939453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,4,1,balanced,0.41510848999023436
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,4,1,balanced,0.7492793273925782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,4,1,balanced,1.0889683532714844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,4,1,balanced,6.206755371093751
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,4,1,balanced,1.1092845153808593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,4,1,balanced,6.824352416992187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,4,1,balanced,1.1835801696777344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,4,1,balanced,1.1070022583007812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,4,1,balanced,1.1455494689941408
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,4,1,balanced,1.131589126586914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,4,1,balanced,1.2021657562255859
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,4,1,balanced,1.1601920318603516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,4,1,balanced,1.1938419342041016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,4,1,balanced,1.1870681762695312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,4,1,balanced,1.2266726684570313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,4,1,balanced,1.2305843353271484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,4,1,balanced,10.400535888671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,4,1,balanced,1.6561677551269534
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,4,1,balanced,1.6062956237792967
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,4,1,balanced,1.7785542297363282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,4,1,balanced,1.8631353759765623
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,4,1,balanced,13.224295654296876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,4,1,balanced,3.2951116943359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,4,1,balanced,3.284038391113281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,4,1,power_law_1.01,0.25786495208740234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,4,1,balanced,4.710285949707031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,4,1,power_law_1.01,0.12818431854248047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,4,1,power_law_1.01,0.40013889312744144
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,4,1,balanced,5.408291015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,4,1,power_law_1.01,1.5107545471191406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,4,1,power_law_1.01,0.6564051055908203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,4,1,power_law_1.01,0.9180204772949219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,4,1,balanced,19.3438330078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,4,1,power_law_1.01,1.7021401977539063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,4,1,balanced,8.199126586914062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,4,1,power_law_1.01,1.749886016845703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,4,1,power_law_1.01,1.6204888916015627
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,4,1,power_law_1.01,1.7847634887695314
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,4,1,power_law_1.01,1.8904389953613283
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,4,1,power_law_1.01,1.9623065185546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,4,1,power_law_1.01,2.080821075439453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,4,1,power_law_1.01,2.353231964111328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,4,1,balanced,9.933140258789063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,4,1,power_law_1.01,2.5318943786621095
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,4,1,power_law_1.01,2.7387738037109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,4,1,power_law_1.01,3.4966296386718754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,4,1,balanced,25.3508837890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,4,1,power_law_1.01,3.359405517578125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,4,1,power_law_1.01,3.5840390014648436
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,4,1,power_law_1.01,4.083403015136719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,4,1,power_law_1.01,4.751585998535156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,4,1,power_law_1.01,0.14864959716796874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,4,1,power_law_1.01,5.83392822265625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,4,1,power_law_1.01,0.3751398468017578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,4,1,power_law_1.01,0.2364467239379883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,4,1,power_law_1.01,0.09341695785522461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,4,1,power_law_1.01,7.314717407226563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,4,1,power_law_1.01,0.504131851196289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,4,1,power_law_1.01,0.8112563323974609
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,4,1,power_law_1.01,0.8745760345458985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,4,1,power_law_1.01,0.9116934204101563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,4,1,power_law_1.01,0.9529542541503906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,4,1,power_law_1.01,0.9711065673828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,4,1,power_law_1.01,0.9918022155761719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,4,1,power_law_1.01,1.018680953979492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,4,1,power_law_1.01,1.0387769317626954
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,4,1,power_law_1.01,1.0864441680908203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,4,1,power_law_1.01,1.1045394897460938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,4,1,power_law_1.01,10.075312499999999
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,4,1,power_law_1.01,1.1376486206054688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,4,1,power_law_1.01,1.2057977294921876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,4,1,power_law_1.01,1.528116455078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,4,1,power_law_1.01,1.6753439331054687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,4,1,power_law_1.01,1.916911926269531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,4,1,power_law_1.01,2.119356231689453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,4,1,power_law_1.01,12.795865478515626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,4,1,power_law_1.01,2.719281921386719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,4,1,power_law_1.2,0.1307808017730713
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,4,1,power_law_1.01,3.3261093139648437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,4,1,power_law_1.2,0.2542848014831543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,4,1,power_law_1.2,0.6039302444458008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,4,1,power_law_1.01,4.612368774414063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,4,1,power_law_1.2,0.8337849426269532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,4,1,power_law_1.2,0.3982809448242187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,4,1,power_law_1.01,5.514886474609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,4,1,power_law_1.2,1.4922221374511717
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,4,1,power_law_1.01,18.21014404296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,4,1,power_law_1.2,1.615608367919922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,4,1,power_law_1.2,1.6764390563964846
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,4,1,power_law_1.2,1.8046258544921876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,4,1,power_law_1.2,1.8395904541015624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,4,1,power_law_1.01,7.835040893554688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,4,1,power_law_1.2,1.919383087158203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,4,1,power_law_1.2,2.030780181884766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,4,1,power_law_1.2,2.158650207519531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,4,1,power_law_1.2,2.51527099609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,4,1,power_law_1.2,2.6845855712890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,4,1,power_law_1.01,10.379349365234376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,4,1,power_law_1.2,2.8771954345703126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,4,1,power_law_1.2,3.450574645996094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,4,1,power_law_1.01,23.854257812500002
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,4,1,power_law_1.2,3.4691595458984374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,4,1,power_law_1.2,3.6826715087890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,4,1,power_law_1.2,4.188118286132813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,4,1,power_law_1.2,0.09512639999389648
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,4,1,power_law_1.2,4.890698852539062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,4,1,power_law_1.2,0.1455635166168213
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,4,1,power_law_1.2,0.24134912490844726
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,4,1,power_law_1.2,0.3423347091674805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,4,1,power_law_1.2,6.086659545898438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,4,1,power_law_1.2,0.8005337524414063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,4,1,power_law_1.2,7.414437866210937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,4,1,power_law_1.2,0.8798719787597656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,4,1,power_law_1.2,0.9046125030517578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,4,1,power_law_1.2,0.9428761291503907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,4,1,power_law_1.2,0.9813574218750001
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,4,1,power_law_1.2,1.0162496185302734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,4,1,power_law_1.2,1.037093734741211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,4,1,power_law_1.2,1.0701900482177735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,4,1,power_law_1.2,10.225984497070312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,4,1,power_law_1.2,1.129401626586914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,4,1,power_law_1.2,1.142080612182617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,4,1,power_law_1.2,0.4563065719604492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,4,1,power_law_1.2,1.1904409790039063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,4,1,power_law_1.2,1.2561293029785157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,4,1,power_law_1.2,1.6066041564941407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,4,1,power_law_1.2,1.6689120483398436
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,4,1,power_law_1.2,1.975465545654297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,4,1,power_law_1.2,2.250832061767578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,4,1,power_law_1.2,13.131429443359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,4,1,power_law_1.2,2.7802227783203124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,1,8,balanced,0.19128576278686524
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,4,1,power_law_1.2,3.3881701660156254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,1,8,balanced,0.19104639053344727
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,1,8,balanced,0.17982336044311525
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,1,8,balanced,0.27676160812377926
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,1,8,balanced,0.27751808166503905
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,1,8,balanced,0.27586175918579103
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,1,8,balanced,0.2770623970031738
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,1,8,balanced,0.2775564765930176
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,1,8,balanced,0.27875839233398436
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,4,1,power_law_1.2,4.334012145996094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,1,8,balanced,0.27864448547363285
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,1,8,balanced,0.2807820892333984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,1,8,balanced,0.30233600616455075
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,1,8,balanced,0.303240966796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,1,8,balanced,0.3159769630432129
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,1,8,balanced,0.3245337677001953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,1,8,balanced,0.33865726470947266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,1,8,balanced,0.35539710998535157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,1,8,balanced,0.41977729797363283
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,1,8,balanced,0.45399295806884765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,1,8,balanced,0.6020390319824218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,1,8,balanced,0.6791001892089843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,1,8,balanced,1.0999616241455077
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,1,8,balanced,1.3992536926269532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,1,8,balanced,2.087080993652344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,4,1,power_law_1.2,5.5554699707031245
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,1,8,balanced,0.09847423553466797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,1,8,balanced,0.09109760284423828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,1,8,balanced,2.7556774902343752
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,1,8,balanced,0.0910591983795166
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,1,8,balanced,0.1278809642791748
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,1,8,balanced,0.1258176040649414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,1,8,balanced,0.12694399833679199
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,1,8,balanced,0.12691967964172363
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,1,8,balanced,0.12719743728637695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,1,8,balanced,0.12773887634277342
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,1,8,balanced,0.12813440322875977
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,1,8,balanced,0.1287820816040039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,1,8,balanced,0.13045887947082518
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,1,8,balanced,0.1304742431640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,1,8,balanced,0.1318617630004883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,1,8,balanced,0.13474047660827637
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,1,8,balanced,0.13671680450439455
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,1,8,balanced,0.1404736042022705
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,1,8,balanced,4.091171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,1,8,balanced,0.1637299156188965
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,1,8,balanced,0.1712063980102539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,1,8,balanced,0.21784832000732424
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,1,8,balanced,0.2922073554992676
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,4,1,power_law_1.2,18.321226806640624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,1,8,balanced,0.3627417755126953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,1,8,balanced,0.5523737716674805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,1,8,balanced,0.7421491241455078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,1,8,balanced,0.22798336029052732
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,1,8,balanced,1.0923840332031252
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,1,8,balanced,5.474299926757813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,1,8,balanced,1.4263591003417968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,1,8,power_law_1.01,0.18031999588012695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,1,8,power_law_1.01,0.27682432174682614
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,1,8,power_law_1.01,0.2758860778808594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,1,8,power_law_1.01,0.2825011253356934
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,1,8,power_law_1.01,0.2763596725463867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,1,8,power_law_1.01,0.306060791015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,1,8,power_law_1.01,0.33048961639404295
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,1,8,power_law_1.01,0.35945343017578124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,4,1,power_law_1.2,7.809528198242188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,1,8,power_law_1.01,0.4253196716308594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,1,8,power_law_1.01,0.27382272720336914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,1,8,power_law_1.01,0.2778188705444336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,1,8,power_law_1.01,0.2939353561401367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,1,8,power_law_1.01,0.6766989135742187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,1,8,power_law_1.01,0.3143398475646973
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,1,8,power_law_1.01,0.9675955200195313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,1,8,power_law_1.01,0.5714995193481445
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,1,8,power_law_1.01,0.3823577499389649
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,1,8,power_law_1.01,1.227804183959961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,1,8,power_law_1.01,1.8737574768066405
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,1,8,power_law_1.01,0.46236415863037106
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,1,8,power_law_1.01,2.3884722900390623
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,1,8,power_law_1.01,3.362471618652344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,1,8,power_law_1.01,0.09251199722290039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,1,8,power_law_1.01,4.849896850585937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,1,8,power_law_1.01,0.13439743995666503
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,1,8,power_law_1.01,0.12883071899414061
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,1,8,power_law_1.01,0.12774271965026857
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,1,8,power_law_1.01,0.1279372787475586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,1,8,power_law_1.01,0.12911487579345704
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,4,1,power_law_1.2,23.9459716796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,1,8,power_law_1.01,8.113558959960937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,4,1,power_law_1.2,10.536522216796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,1,8,power_law_1.01,0.1399948787689209
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,1,8,power_law_1.01,0.14731648445129394
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,1,8,power_law_1.01,0.15678976058959962
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,1,8,power_law_1.01,0.12346879959106445
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,1,8,power_law_1.01,0.1782476806640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,1,8,power_law_1.01,10.281411132812499
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,1,8,power_law_1.01,0.19733888626098633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,1,8,power_law_1.01,0.2586201667785645
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,1,8,power_law_1.01,0.23259519577026366
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,1,8,power_law_1.01,0.3128191947937012
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,1,8,power_law_1.01,0.47992576599121095
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,1,8,power_law_1.01,18.165842285156252
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,1,8,power_law_1.01,0.6418637084960938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,1,8,power_law_1.01,1.2586905670166015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,1,8,power_law_1.01,0.12981760025024414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,1,8,power_law_1.01,0.1318284797668457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,1,8,power_law_1.01,0.13392255783081056
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,1,8,power_law_1.2,0.17470207214355468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,1,8,power_law_1.01,0.1651046371459961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,1,8,power_law_1.2,0.2740633583068848
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,1,8,power_law_1.2,0.2757196807861328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,1,8,power_law_1.01,2.1408320617675782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,1,8,power_law_1.2,0.2890508842468262
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,1,8,power_law_1.01,2.6467276000976563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,1,8,power_law_1.2,0.28542848587036135
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,1,8,power_law_1.2,0.3197580718994141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,1,8,power_law_1.2,0.27381248474121095
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,1,8,power_law_1.2,0.3103513526916504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,1,8,power_law_1.2,0.3284147262573242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,1,8,power_law_1.2,0.33736961364746093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,1,8,power_law_1.2,0.36807296752929686
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,1,8,power_law_1.2,0.38818687438964844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,1,8,power_law_1.2,0.426611213684082
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,1,8,power_law_1.2,0.5088652801513672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,1,8,power_law_1.2,0.27503360748291017
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,1,8,power_law_1.2,0.5744089508056641
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,1,8,power_law_1.2,0.6622911834716797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,1,8,power_law_1.01,0.8742899322509766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,1,8,power_law_1.2,1.336505584716797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,1,8,power_law_1.2,1.065832977294922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,1,8,power_law_1.2,2.06036865234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,1,8,power_law_1.2,2.7136819458007815
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,1,8,power_law_1.2,5.292646484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,1,8,power_law_1.2,3.7466610717773436
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,1,8,power_law_1.2,0.09331583976745605
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,1,8,power_law_1.2,0.12669311523437501
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,1,8,power_law_1.01,4.439822082519531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,1,8,power_law_1.2,0.12623104095458984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,1,8,power_law_1.2,0.139421443939209
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,1,8,power_law_1.2,0.12635135650634766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,1,8,power_law_1.2,0.13525504112243653
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,1,8,power_law_1.2,0.12883968353271485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,1,8,power_law_1.2,0.132294397354126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,1,8,power_law_1.2,0.13195136070251465
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,1,8,power_law_1.2,0.13612031936645508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,1,8,power_law_1.2,0.1377280044555664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,1,8,power_law_1.2,0.15186944007873535
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,1,8,power_law_1.2,0.15784832000732424
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,1,8,power_law_1.2,0.16825855255126954
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,1,8,power_law_1.2,0.1810688018798828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,1,8,power_law_1.2,0.2055398368835449
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,1,8,power_law_1.01,26.94208984375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,1,8,power_law_1.2,0.22495487213134763
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,1,8,power_law_1.2,0.2717452812194824
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,1,8,power_law_1.2,0.3433843231201172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,1,8,power_law_1.2,0.5792512130737305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,1,8,power_law_1.01,5.82105224609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,1,8,power_law_1.2,0.5001663970947265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,1,8,power_law_1.2,1.3793370056152345
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,1,8,power_law_1.2,8.079588012695313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,1,8,power_law_1.2,2.071910400390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,2,4,balanced,0.10017663955688476
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,2,4,balanced,0.0961843204498291
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,2,4,balanced,0.12483967781066893
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,2,4,balanced,0.2002227210998535
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,2,4,balanced,0.19922815322875978
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,2,4,balanced,0.20048639297485354
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,2,4,balanced,0.20053632736206053
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,2,4,balanced,0.20137344360351564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,2,4,balanced,0.19939584732055665
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,2,4,balanced,0.1997862434387207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,2,4,balanced,0.20187776565551757
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,2,4,balanced,0.20437376022338868
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,2,4,balanced,0.20896511077880858
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,2,4,balanced,0.2132236862182617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,2,4,balanced,0.2493824005126953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,2,4,balanced,0.2538521575927734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,2,4,balanced,0.26669567108154296
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,1,8,power_law_1.2,2.6768115234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,2,4,balanced,0.35770111083984374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,2,4,balanced,0.3904537582397461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,2,4,balanced,0.5825215911865234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,2,4,balanced,0.7229261016845703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,2,4,balanced,1.0718067169189454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,1,8,power_law_1.2,13.274161376953126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,2,4,balanced,1.3999244689941406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,2,4,balanced,2.0615679931640623
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,2,4,balanced,2.723285827636719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,2,4,balanced,0.060710401535034174
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,2,4,balanced,0.060439038276672366
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,2,4,balanced,0.07713920116424561
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,2,4,balanced,0.09013119697570801
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,2,4,balanced,0.08994560241699219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,2,4,balanced,0.09039487838745117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,1,8,power_law_1.2,0.9546125030517578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,2,4,balanced,0.09102848052978516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,1,8,power_law_1.2,4.201862487792969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,2,4,balanced,0.10126463890075683
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,2,4,balanced,0.10296575546264648
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,2,4,balanced,0.09836544036865234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,2,4,balanced,4.100429992675781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,2,4,balanced,0.09962240219116211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,2,4,balanced,0.10237183570861816
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,2,4,balanced,0.10159232139587401
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,2,4,balanced,0.10336128234863282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,2,4,balanced,0.10614656448364257
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,2,4,balanced,0.11147647857666017
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,2,4,balanced,0.1377356815338135
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,2,4,balanced,0.13358464241027831
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,2,4,balanced,0.169880313873291
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,2,4,balanced,0.21130111694335935
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,2,4,balanced,0.3079884719848633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,2,4,balanced,0.37810688018798827
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,2,4,balanced,0.5706099319458008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,2,4,balanced,0.09217151641845703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,2,4,balanced,5.463388061523437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,2,4,balanced,0.7301618957519531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,2,4,power_law_1.01,0.09454208374023437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,2,4,power_law_1.01,0.10633600234985352
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,2,4,power_law_1.01,0.19514112472534179
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,2,4,power_law_1.01,0.20068607330322266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,2,4,power_law_1.01,0.19874559402465822
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,2,4,balanced,1.1669427490234374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,2,4,power_law_1.01,0.20100479125976561
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,2,4,power_law_1.01,0.20232831954956057
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,2,4,power_law_1.01,0.20084224700927736
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,2,4,power_law_1.01,0.22027776718139647
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,2,4,balanced,1.5267686462402343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,2,4,power_law_1.01,0.2106150436401367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,2,4,power_law_1.01,0.22401023864746095
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,2,4,power_law_1.01,0.23942144393920897
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,2,4,power_law_1.01,0.2598566436767578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,2,4,power_law_1.01,0.32777473449707034
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,2,4,power_law_1.01,0.3762086486816406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,2,4,power_law_1.01,0.6095667266845703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,2,4,power_law_1.01,0.458287353515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,2,4,power_law_1.01,0.7033074951171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,2,4,power_law_1.01,0.28640384674072267
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,2,4,power_law_1.01,1.1654003143310547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,2,4,power_law_1.01,1.2927565002441406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,2,4,power_law_1.01,1.9724467468261717
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,1,8,power_law_1.2,18.54809814453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,2,4,power_law_1.01,0.05723264217376709
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,2,4,power_law_1.01,2.7761575317382814
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,2,4,power_law_1.01,0.08963583946228028
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,2,4,power_law_1.01,0.09006591796874999
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,2,4,power_law_1.01,0.08997632026672363
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,2,4,power_law_1.01,0.09101696014404297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,2,4,power_law_1.01,4.083674926757813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,2,4,power_law_1.01,0.06465919971466064
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,2,4,power_law_1.01,0.09248384475708007
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,2,4,power_law_1.01,0.09606783866882325
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,2,4,power_law_1.01,0.0947430419921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,2,4,power_law_1.01,0.09598719596862793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,2,4,power_law_1.01,0.09958656311035156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,2,4,power_law_1.01,0.10758912086486816
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,2,4,power_law_1.01,0.11277567863464355
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,2,4,power_law_1.01,0.1209996795654297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,2,4,power_law_1.01,0.13240063667297364
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,2,4,power_law_1.01,0.1375654411315918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,2,4,power_law_1.01,0.1708812713623047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,2,4,power_law_1.01,0.14918784141540525
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,2,4,power_law_1.01,0.2296396827697754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,2,4,power_law_1.01,0.31911935806274416
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,2,4,power_law_1.01,0.4346995162963867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,2,4,power_law_1.01,0.6649037170410156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,2,4,power_law_1.01,5.698804321289062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,1,8,power_law_1.2,5.8393701171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,2,4,power_law_1.01,0.7988390350341797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,2,4,power_law_1.2,0.09601792335510254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,2,4,power_law_1.2,0.11753984451293946
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,2,4,power_law_1.2,0.19720575332641602
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,2,4,power_law_1.2,0.19965696334838867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,2,4,power_law_1.01,1.1244147491455077
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,2,4,power_law_1.2,0.19915008544921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,2,4,power_law_1.2,0.20732288360595702
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,2,4,power_law_1.01,8.706489868164063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,2,4,power_law_1.2,0.20203903198242185
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,2,4,power_law_1.2,0.2038924789428711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,2,4,power_law_1.2,0.20862592697143553
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,2,4,power_law_1.2,0.21474815368652345
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,2,4,power_law_1.2,0.2270732879638672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,2,4,power_law_1.2,0.24769920349121094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,2,4,power_law_1.2,0.2509823989868164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,2,4,power_law_1.2,0.2850022315979004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,2,4,power_law_1.01,1.5196736145019532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,2,4,power_law_1.2,0.34433406829833985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,2,4,power_law_1.2,0.3754867172241211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,2,4,power_law_1.2,0.4486656188964844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,2,4,power_law_1.2,0.639552001953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,2,4,power_law_1.2,0.7938086700439453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,2,4,power_law_1.2,1.2033984375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,2,4,power_law_1.2,2.845313415527344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,2,4,power_law_1.2,1.9730189514160155
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,2,4,power_law_1.01,2.4022412109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,2,4,power_law_1.2,1.4079705810546874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,2,4,power_law_1.2,0.05822591781616211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,2,4,power_law_1.2,0.0639577579498291
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,2,4,power_law_1.2,0.08232704162597657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,2,4,power_law_1.2,0.08704511642456055
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,2,4,power_law_1.2,0.08952704429626465
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,2,4,power_law_1.2,0.09041919708251953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,2,4,power_law_1.2,0.09211775779724121
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,2,4,power_law_1.2,0.09295359611511231
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,2,4,power_law_1.2,0.09478143692016602
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,2,4,power_law_1.2,0.09565823554992677
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,2,4,power_law_1.2,0.0981760025024414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,2,4,power_law_1.2,0.1099187183380127
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,2,4,power_law_1.2,0.11518848419189454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,1,8,power_law_1.2,27.3397314453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,2,4,power_law_1.2,0.12290431976318358
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,2,4,power_law_1.2,0.13156864166259766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,2,4,power_law_1.2,0.13249152183532714
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,2,4,power_law_1.2,0.15378175735473634
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,2,4,power_law_1.2,4.023456115722657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,2,4,power_law_1.2,0.18075904846191407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,2,4,power_law_1.01,15.58892822265625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,2,4,power_law_1.2,0.23116159439086914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,2,4,power_law_1.2,0.4716953659057618
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,2,4,power_law_1.01,3.432469787597656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,2,4,power_law_1.2,0.7639654541015626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,2,4,power_law_1.2,0.3300262451171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,2,4,power_law_1.2,0.5428390502929688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,4,2,balanced,0.0636300802230835
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,4,2,balanced,0.075699200630188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,4,2,balanced,0.10502911567687989
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,4,2,balanced,0.15513471603393553
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,4,2,balanced,0.15422207832336426
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,2,4,power_law_1.2,5.740325317382813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,4,2,balanced,0.1555904006958008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,4,2,balanced,0.1618713569641113
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,4,2,balanced,0.15782784461975097
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,4,2,balanced,0.15890303611755371
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,4,2,balanced,0.15782272338867187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,4,2,balanced,0.15954048156738282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,4,2,balanced,0.16074880599975586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,4,2,balanced,0.1641561508178711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,4,2,balanced,0.16948863983154297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,4,2,balanced,0.2105753517150879
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,4,2,balanced,0.2185932731628418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,2,4,power_law_1.2,1.4969792175292969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,4,2,balanced,0.23052928924560545
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,4,2,balanced,0.3435123062133789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,4,2,balanced,0.407768325805664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,4,2,balanced,0.5770073699951171
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,2,4,power_law_1.2,9.654769897460938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,4,2,balanced,0.7263603210449219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,4,2,balanced,1.0733721923828123
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,4,2,balanced,1.3750796508789063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,4,2,balanced,2.1523878479003904
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,4,2,balanced,0.04925439834594726
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,4,2,balanced,0.05040383815765381
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,4,2,balanced,0.0699609613418579
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,4,2,balanced,2.7438027954101565
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,4,2,balanced,0.07564544200897216
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,4,2,balanced,0.07460864067077637
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,4,2,balanced,0.07720320224761963
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,4,2,balanced,0.07561855792999268
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,4,2,balanced,0.07801472187042237
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,4,2,balanced,0.07696896076202393
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,2,4,power_law_1.2,1.1920166778564454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,4,2,balanced,0.07855743885040282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,4,2,balanced,0.08013312339782715
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,4,2,balanced,0.08754560470581055
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,4,2,balanced,0.08389887809753419
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,4,2,balanced,0.08932607650756837
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,4,2,balanced,0.09706239700317383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,4,2,balanced,0.09538944244384766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,4,2,balanced,0.12562560081481933
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,4,2,balanced,0.13645312309265137
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,4,2,balanced,0.1864793586730957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,4,2,balanced,4.079015808105469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,4,2,balanced,0.22985855102539063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,4,2,balanced,0.33183105468749996
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,4,2,balanced,0.4244249725341797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,4,2,balanced,0.09851776123046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,4,2,balanced,0.624060173034668
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,4,2,power_law_1.01,0.06685696125030519
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,4,2,balanced,0.8427903747558594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,4,2,power_law_1.01,0.07906176090240478
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,4,2,power_law_1.01,0.08834688186645508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,4,2,power_law_1.01,0.15534591674804688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,4,2,power_law_1.01,0.15536767959594727
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,4,2,balanced,5.4504766845703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,4,2,power_law_1.01,0.15801471710205078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,4,2,power_law_1.01,0.1580787181854248
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,4,2,power_law_1.01,0.16020992279052734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,4,2,power_law_1.01,0.1591526412963867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,4,2,balanced,1.625676727294922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,4,2,power_law_1.01,0.16141952514648436
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,4,2,power_law_1.01,0.1703296089172363
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,4,2,power_law_1.01,0.1787174415588379
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,4,2,power_law_1.01,0.21263359069824217
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,4,2,power_law_1.01,0.21002111434936524
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,4,2,power_law_1.01,0.24870399475097654
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,4,2,balanced,1.2382054138183594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,4,2,power_law_1.01,0.26920448303222655
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,4,2,power_law_1.01,0.4451622390747071
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,4,2,power_law_1.01,0.3475545501708984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,4,2,power_law_1.01,0.5343974304199219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,4,2,power_law_1.01,1.443203887939453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,4,2,power_law_1.01,0.8008319854736328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,4,2,power_law_1.01,1.59396484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,4,2,power_law_1.01,1.047531509399414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,4,2,power_law_1.01,0.04153600215911865
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,4,2,power_law_1.01,0.047768321037292484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,4,2,power_law_1.01,0.054090237617492674
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,4,2,power_law_1.01,0.07077888011932373
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,4,2,power_law_1.01,0.07129087924957275
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,4,2,power_law_1.01,0.07416831970214843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,4,2,power_law_1.01,0.07330048084259033
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,4,2,power_law_1.01,0.07582848072052002
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,4,2,power_law_1.01,0.07629183769226075
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,4,2,power_law_1.01,0.07779327869415284
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,4,2,power_law_1.01,0.07937151908874511
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,4,2,power_law_1.01,0.08876543998718261
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,4,2,power_law_1.01,0.09791999816894531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,4,2,power_law_1.01,0.10605567932128905
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,4,2,power_law_1.01,2.6558145141601566
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,4,2,power_law_1.01,0.10931839942932128
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,4,2,power_law_1.01,0.11544320106506348
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,4,2,power_law_1.01,0.10891008377075195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,4,2,power_law_1.01,0.1491609573364258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,4,2,power_law_1.01,0.1815398406982422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,4,2,power_law_1.01,0.2420262336730957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,4,2,power_law_1.01,0.3200435256958008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,2,4,power_law_1.2,2.5077786254882812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,4,2,power_law_1.01,0.4427609634399414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,4,2,power_law_1.01,3.3261669921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,4,2,power_law_1.01,4.726434631347656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,4,2,power_law_1.01,0.6035007858276368
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,4,2,power_law_1.2,0.06517375946044922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,4,2,power_law_1.2,0.0822105598449707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,4,2,power_law_1.2,0.08829567909240724
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,4,2,power_law_1.2,0.1543398380279541
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,4,2,power_law_1.2,0.15037695884704588
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,4,2,power_law_1.2,0.15729408264160155
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,4,2,power_law_1.2,0.1568678379058838
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,4,2,power_law_1.2,0.1577779197692871
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,4,2,power_law_1.2,0.1592473602294922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,4,2,power_law_1.2,0.16126976013183594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,4,2,power_law_1.2,0.16938240051269532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,4,2,power_law_1.01,0.744691162109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,4,2,power_law_1.2,0.17775615692138672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,4,2,power_law_1.2,0.20017663955688478
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,4,2,power_law_1.2,0.2100748825073242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,2,4,power_law_1.2,14.30916748046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,4,2,power_law_1.2,0.24943359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,4,2,power_law_1.2,0.2666444778442383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,4,2,power_law_1.2,0.4525606536865235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,4,2,power_law_1.01,1.3907008361816406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,4,2,power_law_1.2,0.3453721618652344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,4,2,power_law_1.2,0.5288742446899415
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,4,2,power_law_1.2,1.1793113708496095
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,4,2,power_law_1.2,1.474872283935547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,2,4,power_law_1.2,3.362929992675781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,4,2,power_law_1.2,0.7709235382080079
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,4,2,power_law_1.2,2.423057861328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,4,2,power_law_1.01,1.101338882446289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,4,2,power_law_1.2,0.04267007827758789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,4,2,power_law_1.2,0.05178495883941651
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,4,2,power_law_1.2,0.05462528228759765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,4,2,power_law_1.2,0.07276415824890137
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,4,2,power_law_1.2,0.07192192077636719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,4,2,power_law_1.2,0.07553023815155029
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,4,2,power_law_1.2,0.07334400177001953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,4,2,power_law_1.2,0.07651455879211426
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,4,2,power_law_1.2,0.07615615844726562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,4,2,power_law_1.2,0.07794559955596923
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,4,2,power_law_1.2,0.0794764804840088
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,4,2,power_law_1.2,0.09283967971801757
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,4,2,power_law_1.2,0.09796992301940918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,4,2,power_law_1.2,0.10537343978881837
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,4,2,power_law_1.2,2.023937225341797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,4,2,power_law_1.2,0.10980735778808595
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,4,2,power_law_1.2,0.11280896186828612
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,4,2,power_law_1.2,0.14872447967529295
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,4,2,power_law_1.2,0.11547648429870605
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,4,2,power_law_1.2,0.18566783905029297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,4,2,power_law_1.2,0.3355392074584961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,4,2,power_law_1.2,0.23519615173339842
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,4,2,power_law_1.2,0.4274905776977539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,4,2,power_law_1.2,3.799004211425781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,4,2,power_law_1.2,0.5571865463256837
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,4,2,power_law_1.2,1.5226535034179687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,4,2,power_law_1.2,0.7288601684570313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,4,2,power_law_1.01,7.077286376953126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,4,2,power_law_1.01,2.0787149047851563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,8,1,balanced,0.06529280185699463
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,8,1,balanced,0.043724799156188966
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,4,2,power_law_1.2,1.1104000091552735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,8,1,balanced,0.10687295913696289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,8,1,balanced,0.10647232055664062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,4,2,power_law_1.2,6.094970703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,8,1,balanced,0.10544511795043945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,8,1,balanced,0.10769280433654785
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,8,1,balanced,0.1100595188140869
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,8,1,balanced,0.11698752403259279
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,8,1,balanced,0.11151552200317383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,8,1,balanced,0.11226176261901855
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,8,1,balanced,0.11552576065063476
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,8,1,balanced,0.11738623619079588
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,8,1,balanced,0.1332147216796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,8,1,balanced,0.1841164779663086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,8,1,balanced,0.11521792411804199
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,8,1,balanced,0.19065471649169924
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,8,1,balanced,0.2768531227111816
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,8,1,balanced,0.353507194519043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,8,1,balanced,0.5504166412353515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,8,1,balanced,0.6442066955566406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,8,1,balanced,1.0465977478027344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,8,1,balanced,0.12444031715393067
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,8,1,balanced,1.31267333984375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,8,1,balanced,0.17774656295776367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,8,1,balanced,0.06952767848968507
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,8,1,balanced,2.006756439208984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,8,1,balanced,0.17887680053710936
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,8,1,balanced,2.7189996337890627
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,4,2,power_law_1.2,8.064554443359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,8,1,balanced,0.04713791847229004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,8,1,balanced,0.030687360763549804
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,8,1,balanced,0.07547904014587402
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,8,1,balanced,0.07478975772857666
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,8,1,balanced,0.07053311824798583
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,8,1,balanced,0.07647232055664063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,8,1,balanced,0.07798783779144287
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,8,1,balanced,0.07024511814117432
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,8,1,balanced,0.10465087890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,8,1,balanced,0.10713664054870606
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,4,2,power_law_1.2,2.272536315917969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,8,1,balanced,0.13429247856140136
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,8,1,balanced,0.071080322265625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,8,1,balanced,0.18215744018554686
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,8,1,balanced,0.07342400074005126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,8,1,balanced,0.2236140823364258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,8,1,balanced,0.07192512035369873
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,8,1,balanced,0.3305836868286133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,8,1,balanced,0.4161017608642578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,8,1,balanced,0.620654067993164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,8,1,balanced,0.07257984161376954
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,8,1,balanced,0.8079686737060546
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,8,1,balanced,1.2556140899658204
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,8,1,balanced,1.6788677978515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,8,1,power_law_1.01,0.10714943885803223
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,8,1,power_law_1.01,0.10996352195739748
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,8,1,power_law_1.01,0.11357376098632813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,8,1,balanced,0.08431679725646972
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,8,1,balanced,0.10501952171325683
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,8,1,power_law_1.01,0.10967616081237792
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,8,1,power_law_1.01,0.11848640441894531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,8,1,balanced,0.12221759796142577
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,8,1,power_law_1.01,0.1337657642364502
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,8,1,power_law_1.01,0.10817919731140138
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,8,1,power_law_1.01,0.1428102397918701
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,8,1,power_law_1.01,0.06424704074859619
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,8,1,power_law_1.01,0.07427328109741212
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,8,1,power_law_1.01,0.13392127990722655
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,8,1,power_law_1.01,0.17723968505859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,8,1,power_law_1.01,0.17891328811645507
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,8,1,power_law_1.01,0.20565311431884767
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,8,1,power_law_1.01,0.2615103912353516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,8,1,power_law_1.01,0.044337921142578125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,8,1,power_law_1.01,0.3433126449584961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,8,1,power_law_1.01,0.10927167892456055
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,8,1,power_law_1.01,0.14853568077087403
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,8,1,power_law_1.01,0.3852403259277344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,8,1,power_law_1.01,0.7080032348632812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,8,1,power_law_1.01,0.16197439193725588
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,8,1,power_law_1.01,1.0677017974853515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,8,1,power_law_1.01,0.03433343887329102
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,8,1,power_law_1.01,0.046487040519714355
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,8,1,power_law_1.01,1.3713536071777344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,8,1,power_law_1.01,0.05341631889343261
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,8,1,power_law_1.01,0.5399667358398437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,8,1,power_law_1.01,0.07309823989868164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,8,1,power_law_1.01,0.07633600234985352
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,8,1,power_law_1.01,0.08449536323547363
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,8,1,power_law_1.01,0.08989055633544922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,8,1,power_law_1.01,0.06976191997528076
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,8,1,power_law_1.01,0.09726592063903808
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,8,1,power_law_1.01,0.08832575798034667
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,8,1,power_law_1.01,0.10116671562194823
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,8,1,power_law_1.01,0.10820927619934081
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,8,1,power_law_1.01,0.11041343688964844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,8,1,power_law_1.01,0.07049920082092284
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,8,1,power_law_1.01,0.14271936416625977
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,8,1,power_law_1.01,0.15827072143554688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,8,1,power_law_1.01,0.07116032123565674
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,8,1,power_law_1.01,0.2072876739501953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,8,1,power_law_1.01,0.07120575904846191
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,8,1,power_law_1.01,2.0755686950683594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,8,1,power_law_1.01,0.07360127925872803
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,8,1,power_law_1.01,0.2389792060852051
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,8,1,power_law_1.2,0.0441926383972168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,8,1,power_law_1.01,0.33451137542724607
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,8,1,power_law_1.2,0.06984320163726807
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,8,1,power_law_1.01,0.4373612976074218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,8,1,power_law_1.2,0.06292223930358887
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,8,1,power_law_1.2,0.10886400222778321
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,8,1,power_law_1.01,2.6714932250976564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,8,1,power_law_1.01,0.6153887939453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,8,1,power_law_1.2,0.11004480361938476
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,8,1,power_law_1.2,0.10210304260253907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,8,1,power_law_1.2,0.11947392463684081
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,8,1,power_law_1.2,0.1382521629333496
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,8,1,power_law_1.2,0.1476543998718262
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,8,1,power_law_1.2,0.14646528244018556
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,8,1,power_law_1.2,0.10967616081237792
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,8,1,power_law_1.2,0.13304256439208983
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,8,1,power_law_1.2,0.15253184318542481
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,8,1,power_law_1.2,0.18093503952026366
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,8,1,power_law_1.01,0.8387712097167969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,8,1,power_law_1.2,0.11153216361999511
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,8,1,power_law_1.2,0.10874176025390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,8,1,power_law_1.2,0.20171072006225588
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,8,1,power_law_1.2,0.26228416442871094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,8,1,power_law_1.2,0.318209285736084
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,8,1,power_law_1.2,0.3854732894897461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,8,1,power_law_1.01,1.2792658996582031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,8,1,power_law_1.2,0.17763328552246094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,8,1,power_law_1.2,0.5454028701782226
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,8,1,power_law_1.2,0.04689536094665527
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,8,1,power_law_1.2,0.051696639060974124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,8,1,power_law_1.2,0.7286681365966797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,8,1,power_law_1.01,1.7121977233886718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,8,1,power_law_1.2,0.03373568058013916
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,8,1,power_law_1.2,0.07295807838439941
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,8,1,power_law_1.2,0.06970176219940186
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,8,1,power_law_1.2,0.07277376174926757
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,8,1,power_law_1.2,0.07694335937499999
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,8,1,power_law_1.2,0.08610367774963379
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,8,1,power_law_1.2,0.09227904319763183
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,8,1,power_law_1.2,1.041635208129883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,8,1,power_law_1.2,0.06877312183380127
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,8,1,power_law_1.2,0.09541312217712403
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,8,1,power_law_1.2,0.08993599891662599
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,8,1,power_law_1.2,0.09796352386474609
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,8,1,power_law_1.2,0.10904959678649902
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,8,1,power_law_1.2,0.1116639995574951
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,8,1,power_law_1.2,0.14015040397644044
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,8,1,power_law_1.2,0.07092415809631347
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,8,1,power_law_1.2,0.07112959861755372
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,8,1,power_law_1.2,0.15819456100463866
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,8,1,power_law_1.2,0.19852352142333984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,8,1,power_law_1.2,0.23828031539916994
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,8,1,power_law_1.2,1.410946502685547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,8,1,power_law_1.2,0.33591102600097655
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,1,8,balanced,0.23232639312744138
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,1,8,balanced,0.23311231613159178
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,1,8,balanced,0.23181184768676757
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,1,8,balanced,0.3642444610595703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,1,8,balanced,0.36424961090087893
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,1,8,balanced,0.3656332778930664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,8,1,power_law_1.2,0.4370137786865234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,1,8,balanced,0.366748161315918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,1,8,balanced,0.36853118896484377
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,1,8,balanced,0.36967681884765624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,1,8,balanced,0.3741183853149414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,1,8,balanced,0.38188159942626954
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,1,8,balanced,0.4115289688110352
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,1,8,balanced,0.41991424560546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,1,8,balanced,0.4359526443481445
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,1,8,balanced,0.4483468627929687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,1,8,balanced,0.46097023010253907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,1,8,balanced,0.4977996826171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,1,8,balanced,0.5746380615234374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,1,8,balanced,0.6600691223144531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,1,8,balanced,1.0474944305419922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,1,8,balanced,1.2047539520263673
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,1,8,balanced,1.8078746032714843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,8,1,power_law_1.2,0.6304064178466797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,1,8,balanced,2.3667814636230466
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,1,8,balanced,3.469651184082031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,1,8,balanced,4.597380981445313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,1,8,balanced,0.12303744316101076
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,1,8,balanced,0.12064255714416502
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,1,8,balanced,0.12064127922058106
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,1,8,balanced,0.1604185676574707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,1,8,balanced,0.15864576339721678
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,1,8,balanced,0.1591104030609131
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,1,8,balanced,0.15949824333190918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,8,1,power_law_1.2,2.0269375610351563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,1,8,balanced,0.16022783279418945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,1,8,balanced,0.1656768035888672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,1,8,balanced,0.1614784049987793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,1,8,balanced,0.16423168182373046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,1,8,balanced,0.1636992073059082
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,1,8,balanced,0.16460927963256836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,1,8,balanced,7.05705078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,1,8,balanced,0.16844032287597657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,1,8,balanced,0.1753740882873535
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,1,8,balanced,0.1969036865234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,1,8,balanced,0.2077529525756836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,1,8,balanced,0.23750911712646486
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,1,8,balanced,0.2649715232849121
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,1,8,balanced,0.3081267166137695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,1,8,balanced,0.1664908790588379
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,8,1,power_law_1.2,0.8034054565429688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,1,8,balanced,0.47528446197509766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,1,8,balanced,0.6177484893798828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,1,8,balanced,0.83704833984375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,1,8,balanced,1.123916778564453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,1,8,power_law_1.01,0.24783103942871093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,1,8,balanced,1.7992691040039062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,1,8,power_law_1.01,0.36775295257568363
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,1,8,power_law_1.01,0.4108185577392578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,1,8,power_law_1.01,0.36889598846435545
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,1,8,power_law_1.01,0.37288063049316406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,1,8,power_law_1.01,0.37228927612304685
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,1,8,balanced,2.3197785949707033
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,1,8,power_law_1.01,0.39738113403320313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,1,8,power_law_1.01,0.42131713867187504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,1,8,balanced,9.489517822265626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,1,8,power_law_1.01,0.447907829284668
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,1,8,power_law_1.01,0.4475238418579101
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,1,8,power_law_1.01,0.4731609725952149
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,1,8,power_law_1.01,0.5273996734619141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,1,8,power_law_1.01,0.5616255950927734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,1,8,power_law_1.01,0.6038336181640626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,1,8,power_law_1.01,0.7681472015380859
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,1,8,power_law_1.01,0.8844351959228515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,1,8,power_law_1.01,1.2074240112304688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,1,8,power_law_1.01,1.7351795959472658
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,1,8,power_law_1.01,2.2319564819335938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,1,8,power_law_1.01,3.2597760009765624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,1,8,power_law_1.01,4.138133850097656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,8,1,power_law_1.2,2.7191845703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,1,8,power_law_1.01,6.096702880859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,1,8,power_law_1.01,0.12144000053405761
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,8,1,power_law_1.2,1.2460614776611327
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,1,8,power_law_1.01,0.17595008850097654
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,1,8,power_law_1.01,0.16220928192138673
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,1,8,power_law_1.01,0.16059776306152346
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,1,8,power_law_1.01,0.15879039764404296
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,1,8,power_law_1.01,0.16124799728393555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,1,8,power_law_1.01,0.16234495162963866
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,1,8,power_law_1.01,0.16519296646118165
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,1,8,power_law_1.01,0.16615167617797852
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,1,8,power_law_1.01,0.16867584228515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,1,8,power_law_1.01,0.17383552551269532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,1,8,power_law_1.01,0.19600255966186525
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,1,8,power_law_1.01,0.1974732780456543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,1,8,power_law_1.01,0.21207040786743164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,1,8,power_law_1.01,0.24210687637329104
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,1,8,power_law_1.01,0.2502592086791992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,1,8,power_law_1.01,0.2960767936706543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,1,8,power_law_1.01,0.4493376159667969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,1,8,power_law_1.01,8.970723876953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,1,8,power_law_1.01,0.5482726287841797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,1,8,power_law_1.01,0.795758056640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,1,8,power_law_1.01,1.0079859161376954
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,1,8,power_law_1.01,1.6445465087890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,1,8,power_law_1.01,2.2126194763183595
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,8,1,power_law_1.2,1.7455206298828123
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,1,8,power_law_1.01,15.77262939453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,1,8,power_law_1.2,0.24082176208496095
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,1,8,power_law_1.2,0.3669388961791992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,1,8,power_law_1.2,0.36563968658447266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,1,8,power_law_1.2,0.3661849594116211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,1,8,power_law_1.2,0.3682252883911133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,1,8,power_law_1.01,3.983146362304687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,1,8,power_law_1.2,0.371591682434082
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,1,8,power_law_1.2,0.414672622680664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,1,8,power_law_1.2,0.42366718292236333
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,1,8,power_law_1.2,0.4580044937133789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,1,8,power_law_1.2,0.44616577148437503
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,1,8,power_law_1.2,0.4802227020263672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,1,8,power_law_1.2,0.539865608215332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,1,8,power_law_1.2,0.5390476989746094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,1,8,power_law_1.2,0.6244211196899414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,1,8,power_law_1.2,0.7903590393066406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,1,8,power_law_1.2,1.0413990020751953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,1,8,power_law_1.2,1.2166502380371094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,1,8,power_law_1.01,19.83865478515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,1,8,power_law_1.2,1.6765132141113281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,1,8,power_law_1.2,2.1996595764160154
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,1,8,power_law_1.2,3.407468872070312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,1,8,power_law_1.01,3.513680725097656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,1,8,power_law_1.2,4.61931396484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,1,8,power_law_1.01,6.651002807617187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,1,8,power_law_1.2,6.775796508789062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,1,8,power_law_1.2,9.451373901367187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,1,8,power_law_1.01,32.77544921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,1,8,power_law_1.2,0.12092927932739257
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,1,8,power_law_1.2,0.1604377555847168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,1,8,power_law_1.2,0.15835519790649416
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,1,8,power_law_1.2,0.15904640197753905
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,1,8,power_law_1.2,0.16050559997558594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,1,8,power_law_1.2,0.16132991790771484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,1,8,power_law_1.2,0.1633497619628906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,1,8,power_law_1.2,0.16565759658813478
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,1,8,power_law_1.2,0.16818176269531251
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,1,8,power_law_1.2,0.17026176452636718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,1,8,power_law_1.2,0.17501440048217773
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,1,8,power_law_1.2,0.2045913505554199
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,1,8,power_law_1.2,0.1988198471069336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,1,8,power_law_1.2,0.2162688064575195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,1,8,power_law_1.2,0.24524927139282227
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,1,8,power_law_1.2,0.26632448196411135
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,1,8,power_law_1.01,9.3979150390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,1,8,power_law_1.2,0.30605823516845704
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,1,8,power_law_1.2,0.47823745727539063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,1,8,power_law_1.2,0.5623731231689453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,1,8,power_law_1.2,1.057675552368164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,1,8,power_law_1.2,0.8100505828857422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,1,8,power_law_1.2,16.409326171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,1,8,power_law_1.2,2.2558656311035157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,1,8,power_law_1.2,1.7428839111328123
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,1,8,power_law_1.01,45.0821923828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,2,4,balanced,0.12815232276916505
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,2,4,balanced,0.12465408325195312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,2,4,balanced,0.17484031677246095
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,2,4,balanced,0.2846272087097168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,2,4,balanced,0.2824153518676758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,2,4,balanced,0.2889344024658203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,2,4,balanced,0.28915712356567386
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,2,4,balanced,0.2929280090332031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,2,4,balanced,0.2787071990966797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,2,4,balanced,0.2810726356506348
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,2,4,balanced,0.2851110458374023
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,2,4,balanced,0.29832576751708983
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,2,4,balanced,0.31636863708496094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,2,4,balanced,0.31928831100463867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,2,4,balanced,0.3671807861328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,2,4,balanced,0.3714368057250977
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,2,4,balanced,0.3903091049194336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,2,4,balanced,0.5798080062866211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,2,4,balanced,0.6803622436523438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,2,4,balanced,0.973644790649414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,1,8,power_law_1.2,20.139154052734376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,2,4,balanced,1.213824005126953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,1,8,power_law_1.2,3.517082824707031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,2,4,balanced,1.8211546325683592
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,2,4,balanced,2.3326156616210936
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,2,4,balanced,3.4612493896484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,2,4,balanced,4.621578369140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,2,4,balanced,0.0788428783416748
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,2,4,balanced,0.07479167938232421
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,1,8,power_law_1.2,4.297693481445313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,2,4,balanced,7.042169799804688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,2,4,balanced,0.13085824012756347
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,2,4,balanced,0.12355199813842774
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,2,4,balanced,0.12083968162536621
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,2,4,balanced,0.12150783538818359
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,2,4,balanced,0.12191871643066406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,2,4,balanced,0.1228108787536621
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,2,4,balanced,0.12089344024658204
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,2,4,balanced,0.12228863716125489
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,2,4,balanced,0.12386816024780274
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,2,4,balanced,0.1267852783203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,2,4,balanced,0.1283827209472656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,2,4,balanced,0.13237631797790528
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,2,4,balanced,0.1370751953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,2,4,balanced,0.14457216262817382
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,2,4,balanced,0.17960704803466795
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,2,4,balanced,0.18109567642211916
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,2,4,balanced,0.2697983932495117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,2,4,balanced,0.32131839752197267
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,2,4,balanced,0.45669376373291015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,2,4,balanced,0.5775846481323242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,2,4,balanced,9.343983154296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,2,4,balanced,0.8801407623291017
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,2,4,balanced,0.10079615592956544
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,2,4,power_law_1.01,0.12452863693237304
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,2,4,power_law_1.01,0.1458726406097412
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,2,4,power_law_1.01,0.2802739143371582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,2,4,balanced,1.9033445739746093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,2,4,balanced,1.2132364654541017
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,2,4,power_law_1.01,0.286562557220459
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,2,4,power_law_1.01,0.30792831420898437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,2,4,power_law_1.01,0.30021888732910157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,2,4,power_law_1.01,0.2842060852050781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,2,4,balanced,2.3995507812500003
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,2,4,power_law_1.01,0.2921651268005371
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,2,4,power_law_1.01,0.3010995292663574
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,2,4,power_law_1.01,0.3282880020141602
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,2,4,power_law_1.01,0.3492275238037109
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,2,4,power_law_1.01,0.37002494812011716
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,2,4,power_law_1.01,0.4250662231445313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,2,4,power_law_1.01,0.545459213256836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,2,4,power_law_1.01,0.6001676940917969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,2,4,power_law_1.01,0.3590835189819336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,2,4,power_law_1.01,0.7529049682617187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,1,8,power_law_1.2,6.73346435546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,2,4,power_law_1.01,1.041570587158203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,2,4,power_law_1.01,1.3401113891601564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,1,8,power_law_1.2,33.466220703124996
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,2,4,power_law_1.01,1.8412864685058594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,2,4,power_law_1.01,2.5587622070312497
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,2,4,power_law_1.01,4.129714050292969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,2,4,power_law_1.01,0.0749721622467041
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,2,4,power_law_1.01,0.0915494441986084
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,2,4,power_law_1.01,4.816988220214844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,2,4,power_law_1.01,0.1082572841644287
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,2,4,power_law_1.01,0.1225062370300293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,2,4,power_law_1.01,0.11809791564941405
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,2,4,power_law_1.01,0.12470784187316894
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,2,4,power_law_1.01,0.11967359542846681
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,2,4,power_law_1.01,0.12134016036987305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,2,4,power_law_1.01,0.12238592147827147
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,2,4,power_law_1.01,0.12406911849975585
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,2,4,power_law_1.01,0.1270284843444824
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,2,4,power_law_1.01,0.1449625587463379
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,2,4,power_law_1.01,0.14723199844360352
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,2,4,power_law_1.01,0.15648256301879884
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,2,4,power_law_1.01,0.16559999465942382
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,2,4,power_law_1.01,0.17185535430908203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,2,4,power_law_1.01,0.2119897651672363
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,2,4,power_law_1.01,0.29221759796142577
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,2,4,power_law_1.01,0.3399641418457031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,2,4,power_law_1.01,7.380907592773437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,2,4,power_law_1.01,0.5076544189453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,2,4,power_law_1.01,0.6731289672851563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,2,4,power_law_1.01,10.458267822265626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,2,4,power_law_1.01,0.9408102416992188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,2,4,power_law_1.01,1.301739501953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,2,4,power_law_1.01,2.7106164550781253
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,2,4,power_law_1.2,0.13589887619018554
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,2,4,power_law_1.2,0.1500607967376709
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,2,4,power_law_1.01,1.9433229064941409
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,2,4,power_law_1.2,0.2852979278564453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,2,4,power_law_1.2,0.3062220764160156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,2,4,power_law_1.2,0.29123327255249026
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,2,4,power_law_1.2,0.31848064422607425
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,2,4,power_law_1.2,0.28778623580932616
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,2,4,power_law_1.2,0.2886822319030762
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,2,4,power_law_1.2,0.3188390350341797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,2,4,power_law_1.2,0.3343206405639648
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,2,4,power_law_1.2,0.33780609130859374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,2,4,power_law_1.2,0.3640332794189453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,2,4,power_law_1.2,0.36437374114990234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,2,4,power_law_1.2,0.43645183563232426
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,1,8,power_law_1.2,9.412105102539062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,2,4,power_law_1.2,0.5359257507324219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,2,4,power_law_1.01,16.138850097656253
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,2,4,power_law_1.2,0.6453043365478516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,1,8,power_law_1.2,45.74533203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,2,4,power_law_1.2,0.7843392181396485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,2,4,power_law_1.2,1.0127808380126955
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,2,4,power_law_1.2,1.2246361541748048
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,2,4,power_law_1.2,1.9292121887207032
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,2,4,power_law_1.2,2.4455270385742187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,2,4,power_law_1.01,3.823591613769531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,2,4,power_law_1.2,4.15346435546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,2,4,power_law_1.2,0.0745420789718628
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,2,4,power_law_1.2,0.0805196762084961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,2,4,power_law_1.2,0.11035264015197754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,2,4,power_law_1.2,0.11751168251037598
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,2,4,power_law_1.2,0.11715071678161622
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,2,4,power_law_1.2,0.11822463989257812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,2,4,power_law_1.2,0.11929087638854981
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,2,4,power_law_1.2,0.1207756805419922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,2,4,power_law_1.2,0.12305919647216797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,2,4,power_law_1.2,0.12481792449951172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,2,4,power_law_1.2,0.12890624046325683
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,2,4,power_law_1.2,0.149936637878418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,2,4,power_law_1.2,0.1470284843444824
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,2,4,power_law_1.2,0.15858431816101076
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,2,4,power_law_1.2,0.1688729667663574
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,2,4,power_law_1.2,4.814681701660156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,2,4,power_law_1.2,0.1774412727355957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,2,4,power_law_1.2,0.21822080612182618
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,2,4,power_law_1.2,0.30684415817260746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,2,4,power_law_1.2,0.3453926467895508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,2,4,power_law_1.2,0.49074558258056644
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,2,4,power_law_1.2,0.7015436553955079
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,2,4,power_law_1.2,7.505443725585937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,2,4,power_law_1.01,21.490869140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,2,4,power_law_1.01,5.7654528808593755
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,2,4,power_law_1.2,10.489595947265624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,2,4,power_law_1.2,1.3868826293945313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,4,2,balanced,0.08848128318786622
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,2,4,power_law_1.2,1.0019558715820314
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,4,2,balanced,0.10602368354797362
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,4,2,balanced,0.16506624221801758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,4,2,balanced,0.23573631286621094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,4,2,balanced,0.23206527709960936
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,4,2,balanced,0.23748096466064453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,4,2,balanced,0.23443071365356447
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,4,2,balanced,0.23014911651611328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,4,2,balanced,0.22758527755737307
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,4,2,balanced,0.22620288848876954
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,4,2,balanced,0.22835199356079103
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,4,2,balanced,0.23017087936401368
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,4,2,balanced,0.23544191360473632
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,4,2,balanced,0.24172800064086913
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,4,2,balanced,0.36367614746093746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,4,2,balanced,0.3746432113647461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,4,2,balanced,0.38455169677734374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,4,2,balanced,0.5435865783691407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,4,2,balanced,0.6780659484863281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,4,2,balanced,0.9826496124267579
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,4,2,balanced,1.1969843292236328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,2,4,power_law_1.2,1.9752806091308592
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,4,2,balanced,1.8384141540527346
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,4,2,balanced,2.352206115722656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,4,2,balanced,3.5067596435546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,2,4,power_law_1.2,2.736070251464844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,4,2,balanced,0.05434879779815673
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,4,2,balanced,0.06476672172546387
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,4,2,balanced,4.676558227539063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,4,2,balanced,0.0968064022064209
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,4,2,balanced,0.10096768379211425
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,4,2,balanced,0.10126336097717285
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,4,2,balanced,0.10253567695617676
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,4,2,balanced,0.10362879753112793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,4,2,balanced,0.10482815742492675
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,4,2,balanced,0.10571136474609374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,4,2,balanced,0.10733951568603514
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,4,2,balanced,0.11049599647521972
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,4,2,balanced,0.11380736351013183
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,4,2,balanced,0.11740287780761718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,4,2,balanced,0.12479104042053221
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,2,4,power_law_1.2,16.047271728515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,4,2,balanced,0.1303884792327881
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,4,2,balanced,0.13771007537841798
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,4,2,balanced,0.13543423652648925
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,4,2,balanced,7.239261474609376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,4,2,balanced,0.18560895919799805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,4,2,balanced,0.19935615539550783
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,4,2,balanced,0.35743232727050783
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,4,2,balanced,0.49438465118408204
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,4,2,balanced,0.6382463836669923
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,4,2,balanced,0.27134592056274415
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,4,2,balanced,1.036305923461914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,4,2,power_law_1.01,0.09763839721679687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,4,2,balanced,1.2625023651123048
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,4,2,power_law_1.01,0.10901375770568848
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,4,2,power_law_1.01,0.12517503738403318
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,4,2,power_law_1.01,0.22512384414672854
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,4,2,power_law_1.01,0.230645751953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,4,2,power_law_1.01,0.23082111358642576
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,4,2,balanced,1.9387954711914062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,4,2,power_law_1.01,0.23048704147338867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,4,2,power_law_1.01,0.23040639877319338
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,4,2,power_law_1.01,0.24899967193603514
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,4,2,balanced,9.603585205078124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,4,2,power_law_1.01,0.26588287353515627
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,4,2,power_law_1.01,0.2861427116394043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,4,2,power_law_1.01,0.32798721313476564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,4,2,power_law_1.01,0.366693115234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,4,2,balanced,2.6446310424804684
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,4,2,power_law_1.01,0.4292940902709961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,4,2,power_law_1.01,0.47236606597900394
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,4,2,power_law_1.01,0.25558656692504883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,4,2,power_law_1.01,0.5225664138793945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,4,2,power_law_1.01,0.7825536346435547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,4,2,power_law_1.01,0.9831270599365235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,4,2,power_law_1.01,1.2265420532226563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,4,2,power_law_1.01,1.5958834838867186
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,2,4,power_law_1.2,4.0189950561523435
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,4,2,power_law_1.01,2.6457073974609377
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,4,2,power_law_1.01,0.0527027177810669
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,4,2,power_law_1.01,0.06689023971557617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,4,2,power_law_1.01,0.07448832035064698
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,4,2,power_law_1.01,0.10090880393981934
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,4,2,power_law_1.01,0.1014899158477783
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,4,2,power_law_1.01,0.10266624450683595
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,4,2,power_law_1.01,0.10382719993591309
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,4,2,power_law_1.01,0.1051353645324707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,4,2,power_law_1.01,0.10662528038024903
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,4,2,power_law_1.01,0.107774715423584
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,4,2,power_law_1.01,0.1113484764099121
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,4,2,power_law_1.01,0.13240063667297364
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,4,2,power_law_1.01,3.266414184570313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,4,2,power_law_1.01,0.13003135681152345
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,4,2,power_law_1.01,0.1467353630065918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,4,2,power_law_1.01,0.16105600357055666
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,4,2,power_law_1.01,0.17027200698852538
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,4,2,power_law_1.01,0.1710745620727539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,4,2,power_law_1.01,0.2195916748046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,4,2,power_law_1.01,0.24526847839355467
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,4,2,power_law_1.01,0.34026111602783204
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,4,2,power_law_1.01,4.601694641113282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,4,2,power_law_1.01,0.4859904098510742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,4,2,power_law_1.01,6.1567102050781255
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,4,2,power_law_1.01,0.6681369781494141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,4,2,power_law_1.01,0.8696038055419922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,4,2,power_law_1.01,9.155050048828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,4,2,power_law_1.2,0.09255935668945312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,4,2,power_law_1.2,0.10555520057678222
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,4,2,power_law_1.2,0.10731391906738282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,4,2,power_law_1.2,0.22348928451538086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,4,2,power_law_1.2,0.22275583267211915
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,4,2,power_law_1.2,0.22801536560058594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,4,2,power_law_1.2,0.22916223526000978
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,4,2,power_law_1.2,0.23037439346313476
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,4,2,power_law_1.01,1.2171212768554687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,4,2,power_law_1.2,0.24345855712890624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,4,2,power_law_1.2,0.29763839721679686
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,4,2,power_law_1.2,0.26716287612915035
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,4,2,power_law_1.2,0.2999910354614258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,4,2,power_law_1.2,0.31197439193725585
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,4,2,power_law_1.2,0.36384254455566406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,4,2,power_law_1.2,0.42272510528564455
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,4,2,power_law_1.01,1.758217010498047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,4,2,power_law_1.2,0.4588172912597656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,4,2,power_law_1.2,0.5526464080810547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,2,4,power_law_1.2,5.039679870605469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,4,2,power_law_1.2,0.75728515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,4,2,power_law_1.2,0.9651789093017579
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,4,2,power_law_1.2,1.2008882904052736
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,4,2,power_law_1.2,1.8468850708007813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,4,2,power_law_1.01,2.420332794189453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,4,2,power_law_1.2,2.637760009765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,4,2,power_law_1.2,3.278397521972656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,4,2,power_law_1.2,0.054504961967468256
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,4,2,power_law_1.2,0.061619200706481934
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,4,2,power_law_1.2,0.07129216194152832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,2,4,power_law_1.2,24.3823388671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,4,2,power_law_1.2,0.1010547161102295
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,4,2,power_law_1.2,0.10232704162597657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,4,2,power_law_1.2,0.10315008163452148
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,4,2,power_law_1.2,0.1057164764404297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,4,2,power_law_1.2,0.10492799758911134
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,4,2,power_law_1.2,0.10665599822998047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,4,2,power_law_1.2,0.10778240203857421
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,4,2,power_law_1.2,0.11091456413269044
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,4,2,power_law_1.2,0.1364748764038086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,4,2,power_law_1.2,0.1342643165588379
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,4,2,power_law_1.2,0.14484352111816406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,4,2,power_law_1.2,0.15988096237182617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,4,2,power_law_1.2,0.16858495712280272
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,4,2,power_law_1.2,0.16954111099243163
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,4,2,power_law_1.2,0.2167795181274414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,4,2,power_law_1.2,0.24239744186401366
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,4,2,power_law_1.2,4.460405883789063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,4,2,power_law_1.2,0.33963905334472655
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,4,2,power_law_1.01,13.318682861328124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,4,2,power_law_1.2,0.4772864151000976
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,4,2,power_law_1.2,0.6806963348388673
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,4,2,power_law_1.2,5.760946044921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,4,2,power_law_1.01,3.5360818481445313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,4,2,power_law_1.2,0.7719987487792969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,8,1,balanced,0.09812352180480957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,8,1,balanced,0.16869760513305665
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,4,2,power_law_1.2,1.1830976104736328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,8,1,balanced,0.05962175846099853
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,8,1,balanced,0.16724544525146484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,4,2,power_law_1.2,8.447698974609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,8,1,balanced,0.17068735122680664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,8,1,balanced,0.16897344589233398
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,4,2,power_law_1.2,1.8155059814453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,8,1,balanced,0.1724166488647461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,8,1,balanced,0.1731942367553711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,8,1,balanced,0.17368255615234376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,8,1,balanced,0.1987558364868164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,8,1,balanced,0.20168127059936522
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,8,1,balanced,0.20634368896484373
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,8,1,balanced,0.20675584793090823
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,8,1,balanced,0.17558528900146483
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,8,1,balanced,0.17630592346191407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,4,2,power_law_1.2,2.384997100830078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,8,1,balanced,0.2854803276062012
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,8,1,balanced,0.4335174560546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,8,1,balanced,0.1792211151123047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,8,1,balanced,0.5755590438842774
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,8,1,balanced,0.2728070449829102
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,8,1,balanced,0.8396620941162108
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,8,1,balanced,1.1290924835205078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,8,1,balanced,1.6402053833007812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,8,1,balanced,0.18175872802734375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,8,1,balanced,2.22765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,8,1,balanced,0.10236736297607422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,8,1,balanced,0.10147199630737305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,8,1,balanced,0.10231871604919433
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,8,1,balanced,0.1032249641418457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,8,1,balanced,0.10358976364135741
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,8,1,balanced,3.320138854980469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,8,1,balanced,0.10468544006347655
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,8,1,balanced,0.10761407852172851
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,8,1,balanced,0.10774208068847657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,8,1,balanced,0.1105459213256836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,8,1,balanced,0.1120531177520752
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,8,1,balanced,0.1184671974182129
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,8,1,balanced,0.14717439651489256
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,8,1,balanced,0.15025343894958496
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,8,1,balanced,0.13416640281677245
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,8,1,balanced,0.16970815658569335
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,8,1,balanced,0.1832499122619629
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,8,1,balanced,0.2566265678405762
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,8,1,balanced,0.3322489547729492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,8,1,balanced,0.4772806549072266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,8,1,balanced,4.3788122558593745
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,8,1,balanced,0.6433318328857422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,8,1,balanced,0.10141695976257323
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,8,1,balanced,0.9727436828613282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,8,1,balanced,1.2855783081054688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,8,1,balanced,1.9654246520996093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,8,1,power_law_1.01,0.12122048377990721
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,4,2,power_law_1.2,11.87052490234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,8,1,balanced,2.587632751464844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,8,1,balanced,0.04620736122131348
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,8,1,power_law_1.01,0.09177599906921387
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,8,1,power_law_1.01,0.17112960815429687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,8,1,power_law_1.01,0.1718169593811035
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,8,1,power_law_1.01,0.06016448020935059
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,8,1,power_law_1.01,0.1792192077636719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,8,1,power_law_1.01,0.16373952865600586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,8,1,balanced,0.06657855987548829
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,8,1,power_law_1.01,0.16964160919189453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,8,1,power_law_1.01,0.18146175384521485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,8,1,power_law_1.01,0.2504518318176269
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,8,1,power_law_1.01,0.24877695083618162
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,8,1,power_law_1.01,0.2586604881286621
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,8,1,power_law_1.01,0.2647769546508789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,8,1,power_law_1.01,0.3688441467285156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,4,2,power_law_1.2,3.2596658325195316
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,8,1,power_law_1.01,0.3719468688964844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,8,1,power_law_1.01,0.1722368049621582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,8,1,power_law_1.01,0.5153971099853516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,8,1,power_law_1.01,0.19106752395629883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,8,1,power_law_1.01,0.18935680389404297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,8,1,power_law_1.01,0.1916806411743164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,8,1,power_law_1.01,0.6325183868408203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,8,1,power_law_1.01,0.9067596435546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,8,1,power_law_1.01,0.07375743865966797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,8,1,power_law_1.01,0.09986880302429199
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,8,1,power_law_1.01,0.100315523147583
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,8,1,power_law_1.01,0.10121919631958007
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,8,1,power_law_1.01,1.2033977508544922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,8,1,power_law_1.01,0.10299776077270509
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,8,1,power_law_1.01,0.10415167808532715
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,8,1,power_law_1.01,0.10303808212280274
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,8,1,power_law_1.01,0.10761280059814453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,8,1,power_law_1.01,0.11345024108886718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,8,1,power_law_1.01,0.1149830436706543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,8,1,power_law_1.01,0.11734208106994629
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,8,1,power_law_1.01,0.13649727821350097
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,8,1,power_law_1.01,0.14484479904174805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,8,1,power_law_1.01,0.15471039772033693
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,8,1,power_law_1.01,0.15876928329467774
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,8,1,power_law_1.01,0.21174655914306642
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,8,1,power_law_1.01,1.781425323486328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,8,1,power_law_1.01,0.2190719985961914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,8,1,power_law_1.01,0.2993657684326172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,8,1,power_law_1.01,0.3719046401977539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,8,1,power_law_1.01,0.5133388900756837
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,8,1,power_law_1.01,2.292869110107422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,8,1,power_law_1.01,0.04592832088470459
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,8,1,power_law_1.01,0.681712646484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,8,1,power_law_1.2,0.060374398231506345
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,8,1,power_law_1.01,0.0643398380279541
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,8,1,power_law_1.01,0.969324188232422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,8,1,power_law_1.2,0.16804479598999023
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,8,1,power_law_1.2,0.08985088348388673
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,8,1,power_law_1.01,3.3700628662109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,8,1,power_law_1.2,0.17758848190307616
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,8,1,power_law_1.2,0.1723468780517578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,8,1,power_law_1.2,0.17616064071655274
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,8,1,power_law_1.2,0.17929599761962892
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,8,1,power_law_1.2,0.10087295532226563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,8,1,power_law_1.01,1.3217701721191406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,8,1,power_law_1.2,0.19179519653320312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,8,1,power_law_1.2,0.1952390480041504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,8,1,power_law_1.2,0.2509523200988769
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,8,1,power_law_1.2,0.2531987190246582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,8,1,power_law_1.2,0.2632070350646972
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,8,1,power_law_1.2,0.1631328010559082
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,8,1,power_law_1.2,0.26617343902587887
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,8,1,power_law_1.2,0.34819393157958983
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,8,1,power_law_1.2,0.18593408584594726
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,8,1,power_law_1.2,0.36996223449707033
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,8,1,power_law_1.2,0.5103020858764649
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,8,1,power_law_1.2,0.19304000854492187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,8,1,power_law_1.01,4.475519409179688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,8,1,power_law_1.2,0.6351264190673829
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,8,1,power_law_1.01,2.0095974731445314
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,8,1,power_law_1.2,0.938625259399414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,8,1,power_law_1.2,1.132053756713867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,8,1,power_law_1.01,2.6088916015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,8,1,power_law_1.2,0.10054911613464354
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,8,1,power_law_1.2,0.09954560279846192
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,8,1,power_law_1.2,0.10221695899963379
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,8,1,power_law_1.2,0.10349504470825195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,8,1,power_law_1.2,0.10266559600830079
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,8,1,power_law_1.2,0.10533184051513671
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,8,1,power_law_1.2,0.06640639781951904
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,8,1,power_law_1.2,0.10791232109069823
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,8,1,power_law_1.2,0.11347071647644043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,8,1,power_law_1.2,0.11650495529174805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,8,1,power_law_1.2,0.11643903732299805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,8,1,power_law_1.2,0.1359116840362549
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,8,1,power_law_1.2,0.04644608020782471
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,8,1,power_law_1.2,0.157903995513916
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,8,1,power_law_1.2,0.15977215766906738
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,8,1,power_law_1.2,1.7840173339843752
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,8,1,power_law_1.2,0.20225664138793947
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,8,1,power_law_1.2,0.220449275970459
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,8,1,power_law_1.2,0.29577600479125976
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,8,1,power_law_1.2,0.06359039783477784
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,8,1,power_law_1.2,0.37298431396484377
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,8,1,power_law_1.2,0.14476672172546387
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,8,1,power_law_1.2,2.276769866943359
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,8,1,power_law_1.2,0.5182227325439454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,1,8,balanced,0.07155072212219238
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,1,8,balanced,0.08099200248718261
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,1,8,balanced,0.11033472061157226
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,1,8,balanced,0.2043391990661621
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,1,8,balanced,0.42645889282226557
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,8,1,power_law_1.2,0.6812313842773438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,1,8,balanced,0.7099890899658203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,1,8,balanced,1.0473535919189454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,1,8,balanced,1.0544461059570314
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,1,8,balanced,1.0595340728759766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,1,8,balanced,1.060226593017578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,1,8,balanced,1.0578189086914063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,1,8,balanced,1.0637529754638673
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,1,8,balanced,1.071335678100586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,1,8,balanced,1.0826252746582032
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,8,1,power_law_1.2,1.041866912841797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,1,8,balanced,1.1054246520996094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,1,8,balanced,1.1188889312744141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,1,8,balanced,1.1491609954833986
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,1,8,balanced,1.209016342163086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,1,8,balanced,1.274456329345703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,1,8,balanced,1.381913604736328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,1,8,balanced,1.4752575683593752
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,1,8,balanced,1.7360231018066408
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,1,8,balanced,1.9639935302734375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,8,1,power_law_1.2,3.352825012207031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,1,8,balanced,2.7956890869140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,1,8,balanced,0.045099520683288576
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,1,8,balanced,0.05452288150787353
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,1,8,balanced,0.06915711879730224
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,1,8,balanced,3.5318399047851563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,1,8,balanced,0.10973952293395997
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,1,8,balanced,0.19578752517700196
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,1,8,balanced,0.35819263458251954
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,1,8,balanced,0.3904064178466797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,1,8,balanced,0.3934463882446289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,8,1,power_law_1.2,1.3307948303222656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,1,8,balanced,0.3965363311767578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,1,8,balanced,0.39913726806640626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,1,8,balanced,0.39375614166259765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,1,8,balanced,0.3939763259887695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,1,8,balanced,0.3966195297241211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,1,8,balanced,5.301757202148438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,1,8,balanced,0.40402175903320314
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,1,8,balanced,0.41228542327880857
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,1,8,balanced,0.4191846466064453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,1,8,balanced,0.44541183471679685
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,1,8,balanced,0.47499263763427735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,1,8,balanced,0.5057587051391602
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,1,8,balanced,0.5587686538696289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,1,8,balanced,0.6166553497314453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,1,8,balanced,0.7445977783203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,1,8,balanced,0.8899404907226562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,1,8,balanced,6.874945068359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,1,8,balanced,1.2510758209228516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,1,8,power_law_1.01,0.21021951675415038
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,1,8,balanced,1.5151679992675782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,1,8,power_law_1.01,0.42506881713867184
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,1,8,power_law_1.01,0.4529420852661133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,1,8,power_law_1.01,0.4640665435791016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,1,8,balanced,2.1512754821777342
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,1,8,power_law_1.01,0.5951961517333985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,1,8,power_law_1.01,0.5380070495605469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,1,8,balanced,2.7943231201171876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,1,8,power_law_1.01,0.9058879852294922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,1,8,power_law_1.01,0.8760358428955077
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,1,8,power_law_1.01,0.8275328063964844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,1,8,power_law_1.01,0.8399001312255858
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,1,8,power_law_1.01,0.9027136230468751
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,1,8,power_law_1.01,0.906055679321289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,1,8,power_law_1.01,0.9194329833984375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,1,8,power_law_1.01,0.956552963256836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,1,8,power_law_1.01,0.9488934326171876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,1,8,power_law_1.01,1.0654630279541015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,1,8,power_law_1.01,1.0938585662841798
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,1,8,power_law_1.01,1.2383833312988282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,1,8,power_law_1.01,1.2922329711914062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,1,8,power_law_1.01,1.4718028259277343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,8,1,power_law_1.2,4.47197509765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,1,8,power_law_1.01,1.7946444702148436
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,1,8,power_law_1.01,2.2837286376953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,1,8,power_law_1.01,0.107774715423584
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,1,8,power_law_1.01,2.703827209472656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,1,8,power_law_1.01,0.19117439270019532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,1,8,power_law_1.01,0.23969024658203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,1,8,power_law_1.01,0.2414041519165039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,1,8,power_law_1.01,0.285295352935791
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,1,8,power_law_1.01,3.8229043579101565
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,1,8,power_law_1.01,0.2760793685913086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,1,8,power_law_1.01,0.3284147262573242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,1,8,power_law_1.01,0.3561868667602539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,1,8,power_law_1.01,0.34149375915527347
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,8,1,power_law_1.2,1.9264857482910156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,1,8,power_law_1.01,0.35299713134765626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,1,8,power_law_1.01,0.3633203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,1,8,power_law_1.01,0.36209022521972656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,1,8,power_law_1.01,0.36612735748291014
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,1,8,power_law_1.01,5.045277404785156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,1,8,power_law_1.01,0.3700390243530273
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,1,8,power_law_1.01,0.3674534225463867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,1,8,power_law_1.01,0.388930549621582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,1,8,power_law_1.01,0.40302337646484376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,1,8,power_law_1.01,0.44605182647705083
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,1,8,power_law_1.01,0.4763967895507813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,1,8,power_law_1.01,0.5562265777587891
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,1,8,power_law_1.01,0.6303142547607422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,1,8,power_law_1.01,0.8209625244140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,1,8,power_law_1.01,0.9243904113769531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,1,8,power_law_1.01,1.3856857299804688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,1,8,power_law_1.2,0.20397567749023438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,1,8,power_law_1.01,1.887906494140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,1,8,power_law_1.01,8.402672729492187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,1,8,power_law_1.2,0.4234444808959961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,1,8,power_law_1.2,0.399791374206543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,1,8,power_law_1.01,2.5262860107421874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,1,8,power_law_1.2,0.46480766296386716
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,1,8,power_law_1.2,0.5135449600219727
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,1,8,power_law_1.2,0.6159219360351562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,1,8,power_law_1.01,9.331526489257813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,1,8,power_law_1.2,0.9038848114013671
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,1,8,power_law_1.01,3.4003790283203124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,1,8,power_law_1.2,0.8717670440673828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,1,8,power_law_1.2,0.865301742553711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,1,8,power_law_1.2,0.852152328491211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,1,8,power_law_1.2,0.8691212463378907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,1,8,power_law_1.2,0.9059737396240234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,1,8,power_law_1.2,0.9960435485839844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,1,8,power_law_1.2,1.017902069091797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,1,8,power_law_1.2,1.096236801147461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,1,8,power_law_1.2,0.8322611236572266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,1,8,power_law_1.2,1.1666790771484377
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,1,8,power_law_1.2,1.2184832000732422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,1,8,power_law_1.2,1.4017677307128906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,1,8,power_law_1.2,1.6027801513671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,1,8,power_law_1.2,2.047810516357422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,1,8,power_law_1.2,2.3693862915039063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,1,8,power_law_1.2,0.11118592262268065
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,1,8,power_law_1.2,0.1891430473327637
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,1,8,power_law_1.2,0.1756505584716797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,1,8,power_law_1.2,3.153509216308594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,1,8,power_law_1.2,0.24371711730957032
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,1,8,power_law_1.2,0.26252672195434573
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,1,8,power_law_1.2,0.30090240478515623
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,1,8,power_law_1.2,0.3088307189941406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,1,8,power_law_1.2,0.3334502410888672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,1,8,power_law_1.2,0.32413440704345703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,1,8,power_law_1.2,0.3459584045410156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,1,8,power_law_1.2,0.35967105865478516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,1,8,power_law_1.2,4.815693969726563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,1,8,power_law_1.2,0.36286975860595705
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,1,8,power_law_1.2,0.3506662368774414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,1,8,power_law_1.2,0.3715366363525391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,1,8,power_law_1.2,0.3764518356323242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,1,8,power_law_1.2,0.3924313735961914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,1,8,power_law_1.2,0.4246745681762695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,1,8,power_law_1.2,0.4718016052246094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,1,8,power_law_1.2,5.585938110351562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,1,8,power_law_1.2,0.5116454315185547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,1,8,power_law_1.2,0.6072985458374023
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,8,1,power_law_1.2,2.6340231323242187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,1,8,power_law_1.2,0.6682994842529297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,1,8,power_law_1.2,0.8817727661132813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,1,8,power_law_1.2,1.0812493133544923
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,2,4,balanced,0.06681983947753907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,2,4,balanced,0.08009087562561035
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,2,4,balanced,0.11830400466918944
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,2,4,balanced,0.18820863723754883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,1,8,power_law_1.2,1.7618009948730466
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,2,4,balanced,0.41838718414306636
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,1,8,power_law_1.2,2.170163269042969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,2,4,balanced,0.6960063934326172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,1,8,power_law_1.2,8.257241821289062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,2,4,balanced,0.6997837066650391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,2,4,balanced,0.6985215759277343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,1,8,power_law_1.2,2.847832336425781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,2,4,balanced,1.0107071685791016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,2,4,balanced,1.010760955810547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,2,4,balanced,1.028505630493164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,2,4,balanced,1.0334297943115236
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,2,4,balanced,1.03328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,2,4,balanced,1.0482879638671876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,1,8,power_law_1.2,4.141139221191406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,2,4,balanced,1.066513900756836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,2,4,balanced,1.075257568359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,2,4,balanced,1.1003404998779298
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,2,4,balanced,1.1618994903564452
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,2,4,balanced,1.2372096252441405
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,2,4,balanced,1.3569778442382812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,2,4,balanced,1.4585382080078124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,2,4,balanced,1.7394752502441406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,2,4,balanced,0.04158336162567139
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,2,4,balanced,1.9342247009277345
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,2,4,balanced,0.05284863948822022
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,2,4,balanced,0.06921728134155274
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,2,4,balanced,0.11145343780517578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,2,4,balanced,0.1931161689758301
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,2,4,balanced,0.35355777740478517
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,1,8,power_law_1.2,13.04912353515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,2,4,balanced,0.35949695587158204
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,2,4,balanced,2.7210842895507814
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,2,4,balanced,0.36189056396484376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,2,4,balanced,0.3884262466430664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,2,4,balanced,0.40085887908935547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,2,4,balanced,0.39182975769042966
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,2,4,balanced,0.39746814727783203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,2,4,balanced,0.3976755142211914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,2,4,balanced,3.451404724121094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,2,4,balanced,0.40487422943115237
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,2,4,balanced,0.4126003265380859
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,2,4,balanced,0.4187187194824219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,2,4,balanced,0.43559936523437504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,2,4,balanced,0.46814334869384766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,2,4,balanced,0.4864806365966797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,2,4,balanced,0.5419801712036133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,2,4,balanced,0.6009996795654298
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,2,4,balanced,0.6993548583984375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,2,4,balanced,0.831942367553711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,2,4,balanced,5.116392822265626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,2,4,power_law_1.01,0.11819904327392579
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,2,4,balanced,1.1838489532470704
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,2,4,power_law_1.01,0.18417280197143554
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,2,4,balanced,1.4098240661621095
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,2,4,power_law_1.01,0.2591449546813965
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,2,4,power_law_1.01,0.32841728210449217
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,2,4,power_law_1.01,0.44370559692382816
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,2,4,power_law_1.01,0.45768577575683594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,2,4,balanced,2.0300889587402344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,2,4,power_law_1.01,0.6211238479614257
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,2,4,power_law_1.01,0.6148108673095704
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,2,4,balanced,6.420633544921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,2,4,power_law_1.01,0.8363942718505859
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,2,4,power_law_1.01,0.8587251281738281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,2,4,balanced,2.6762200927734376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,2,4,power_law_1.01,0.8244889831542969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,2,4,power_law_1.01,0.8870937347412109
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,2,4,power_law_1.01,0.8486860656738282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,2,4,power_law_1.01,0.8790617370605469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,2,4,power_law_1.01,0.9380761718749999
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,2,4,power_law_1.01,0.9528768157958984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,2,4,power_law_1.01,1.0235289764404296
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,2,4,power_law_1.01,1.1837964630126954
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,2,4,power_law_1.01,1.2255859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,2,4,power_law_1.01,1.4415245056152344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,2,4,power_law_1.01,1.5969343566894532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,2,4,power_law_1.01,0.06734335899353028
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,2,4,power_law_1.01,0.10941184043884278
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,2,4,power_law_1.01,2.074401245117188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,2,4,power_law_1.01,0.1429145622253418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,2,4,power_law_1.01,0.16236543655395508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,2,4,power_law_1.01,2.3705062866210938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,2,4,power_law_1.01,0.2293120002746582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,2,4,power_law_1.01,0.22951679229736327
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,2,4,power_law_1.01,0.29885824203491207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,2,4,power_law_1.01,0.319944953918457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,2,4,power_law_1.01,0.3320115280151367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,2,4,power_law_1.01,0.34558078765869144
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,2,4,power_law_1.01,0.35032447814941403
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,2,4,power_law_1.01,0.36049407958984375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,2,4,power_law_1.01,3.476870422363281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,2,4,power_law_1.01,0.3511283111572266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,2,4,power_law_1.01,0.3601356887817383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,2,4,power_law_1.01,0.37421825408935544
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,2,4,power_law_1.01,0.3891878509521484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,2,4,power_law_1.01,0.39605888366699216
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,2,4,power_law_1.01,0.4439667129516602
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,2,4,power_law_1.01,0.4759052658081055
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,2,4,power_law_1.01,4.337556457519531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,2,4,power_law_1.01,0.557212142944336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,2,4,power_law_1.01,0.6214937591552735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,2,4,power_law_1.01,0.7803839874267579
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,2,4,power_law_1.01,0.926723861694336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,2,4,power_law_1.2,0.11690496444702148
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,2,4,power_law_1.2,0.18360576629638672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,2,4,power_law_1.2,0.21213184356689455
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,2,4,power_law_1.01,1.3110488891601562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,2,4,power_law_1.2,0.3126451110839844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,2,4,power_law_1.01,1.722123565673828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,2,4,power_law_1.2,0.43418624877929685
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,2,4,power_law_1.2,0.43234943389892583
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,2,4,power_law_1.01,6.24956787109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,2,4,power_law_1.2,0.5730521774291992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,2,4,power_law_1.2,0.5586713409423828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,2,4,power_law_1.01,2.437626953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,2,4,power_law_1.2,0.8743885040283204
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,2,4,power_law_1.2,0.8352870178222656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,2,4,power_law_1.2,0.8596825408935548
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,2,4,power_law_1.01,2.989726867675781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,2,4,power_law_1.2,0.872069091796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,2,4,power_law_1.01,7.917576904296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,2,4,power_law_1.2,0.923125762939453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,2,4,power_law_1.2,0.9900927734374999
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,2,4,power_law_1.2,0.9861325073242189
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,2,4,power_law_1.2,1.0800640106201171
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,2,4,power_law_1.2,1.181743392944336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,2,4,power_law_1.2,1.2760832214355469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,2,4,power_law_1.2,0.8697779083251953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,2,4,power_law_1.2,1.5237338256835937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,2,4,power_law_1.2,1.7511500549316406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,2,4,power_law_1.2,0.06860544204711913
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,2,4,power_law_1.2,0.10930432319641112
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,2,4,power_law_1.2,0.12476672172546385
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,2,4,power_law_1.2,2.2396978759765624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,2,4,power_law_1.2,0.1619443130493164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,2,4,power_law_1.2,0.17993600845336916
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,2,4,power_law_1.2,2.6475009155273437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,2,4,power_law_1.2,0.24524288177490233
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,2,4,power_law_1.2,0.27843967437744144
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,2,4,power_law_1.2,0.29252607345581055
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,2,4,power_law_1.2,0.3286630249023438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,2,4,power_law_1.2,0.324403190612793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,2,4,power_law_1.2,3.5015386962890624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,2,4,power_law_1.2,0.34948608398437503
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,2,4,power_law_1.2,0.34979969024658203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,2,4,power_law_1.2,0.35750526428222656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,2,4,power_law_1.2,0.36073726654052735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,2,4,power_law_1.2,0.38563072204589843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,2,4,power_law_1.2,0.3908544158935547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,2,4,power_law_1.2,0.40950271606445315
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,2,4,power_law_1.2,0.46102272033691405
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,2,4,power_law_1.2,0.4900390243530273
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,2,4,power_law_1.2,4.86166015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,2,4,power_law_1.2,0.5839295959472656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,2,4,power_law_1.2,0.6708914947509765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,2,4,power_law_1.2,0.8382822418212891
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,2,4,power_law_1.2,0.9869977569580077
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,4,2,balanced,0.06370816230773926
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,4,2,balanced,0.08023679733276368
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,4,2,balanced,0.10902400016784668
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,2,4,power_law_1.2,1.3652621459960936
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,4,2,balanced,0.1854694366455078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,2,4,power_law_1.2,6.627142333984375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,4,2,balanced,0.40697982788085935
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,2,4,power_law_1.2,1.7120127868652344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,4,2,balanced,0.6930931091308594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,4,2,balanced,0.6984934234619141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,4,2,balanced,0.6982195281982422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,4,2,balanced,0.7034368133544922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,4,2,balanced,0.7116659545898438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,2,4,power_law_1.2,2.78210693359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,4,2,balanced,1.006704635620117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,4,2,balanced,1.0098047637939453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,4,2,balanced,1.0241241455078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,2,4,power_law_1.2,9.040878295898438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,4,2,balanced,1.046719970703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,2,4,power_law_1.2,3.467727355957031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,4,2,balanced,1.0589849853515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,4,2,balanced,1.0787366485595702
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,4,2,balanced,0.6917644500732422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,4,2,balanced,1.129638366699219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,4,2,balanced,1.1948774719238282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,4,2,balanced,1.3197914123535157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,4,2,balanced,0.041368322372436525
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,4,2,balanced,1.432500457763672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,4,2,balanced,0.054606080055236816
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,4,2,balanced,0.07171199798583985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,4,2,balanced,1.7664883422851563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,4,2,balanced,0.11507840156555176
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,4,2,balanced,0.2053388786315918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,4,2,balanced,2.012953643798828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,4,2,balanced,0.3801190567016602
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,4,2,balanced,0.3814681625366211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,4,2,balanced,0.38246273040771483
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,4,2,balanced,0.3881702423095703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,4,2,balanced,0.3913420867919922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,4,2,balanced,2.747185974121094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,4,2,balanced,0.3936422348022461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,4,2,balanced,0.437259521484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,4,2,balanced,0.44104961395263675
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,4,2,balanced,0.44776832580566406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,4,2,balanced,0.45766273498535154
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,4,2,balanced,0.45879169464111325
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,4,2,balanced,3.4648333740234376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,4,2,balanced,0.4675699234008789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,4,2,balanced,0.49209854125976565
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,4,2,balanced,0.5173644638061523
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,4,2,balanced,0.5627302551269532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,4,2,balanced,0.6146329498291015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,4,2,balanced,0.722108154296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,4,2,balanced,0.8223513793945312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,4,2,power_law_1.01,0.07988607883453369
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,4,2,power_law_1.01,0.10824319839477539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,4,2,balanced,1.2402342224121095
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,4,2,balanced,5.098377075195312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,4,2,power_law_1.01,0.14765567779541017
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,4,2,power_law_1.01,0.20569984436035157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,4,2,balanced,1.4459187316894533
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,4,2,power_law_1.01,0.3575961685180664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,4,2,power_law_1.01,0.4189260864257813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,4,2,power_law_1.01,0.5010508728027344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,4,2,power_law_1.01,0.5407487869262695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,4,2,balanced,2.078577880859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,4,2,power_law_1.01,0.5584691238403321
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,4,2,balanced,6.4365478515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,4,2,power_law_1.01,0.6034880065917969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,4,2,power_law_1.01,0.8076172637939454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,4,2,balanced,2.720531311035156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,4,2,power_law_1.01,0.8251686096191406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,4,2,power_law_1.01,0.8874368286132812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,4,2,power_law_1.01,0.935266571044922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,4,2,power_law_1.01,0.5596748733520508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,4,2,power_law_1.01,0.9463168334960936
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,4,2,power_law_1.01,1.0896217346191406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,4,2,power_law_1.01,1.2053798675537108
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,4,2,power_law_1.01,1.3351129150390624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,4,2,power_law_1.01,0.982519073486328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,4,2,power_law_1.01,1.5114559936523437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,4,2,power_law_1.01,1.8312690734863282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,4,2,power_law_1.01,0.07206016063690186
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,4,2,power_law_1.01,2.265086669921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,4,2,power_law_1.01,0.09358336448669433
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,4,2,power_law_1.01,0.13757951736450197
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,4,2,power_law_1.01,0.0531763219833374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,4,2,power_law_1.01,3.0405758666992186
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,4,2,power_law_1.01,0.17885824203491213
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,4,2,power_law_1.01,0.22724735260009768
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,4,2,power_law_1.01,0.2875391960144043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,4,2,power_law_1.01,0.3110003280639648
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,4,2,power_law_1.01,0.31179008483886717
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,4,2,power_law_1.01,0.3227264022827149
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,4,2,power_law_1.01,0.33731327056884763
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,4,2,power_law_1.01,0.3868288040161133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,4,2,power_law_1.01,0.3987315368652344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,4,2,power_law_1.01,0.4151180648803711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,4,2,power_law_1.01,0.41313022613525396
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,4,2,power_law_1.01,0.43725311279296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,4,2,power_law_1.01,0.46213119506835937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,4,2,power_law_1.01,0.51325439453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,4,2,power_law_1.01,0.5840179061889648
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,4,2,power_law_1.01,3.9244927978515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,4,2,power_law_1.01,0.6515634918212891
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,4,2,power_law_1.01,5.633385009765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,4,2,power_law_1.01,0.7990962982177734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,4,2,power_law_1.01,0.9748070526123047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,4,2,power_law_1.2,0.07956096172332763
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,4,2,power_law_1.01,0.35790592193603515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,4,2,power_law_1.2,0.10947711944580078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,4,2,power_law_1.2,0.1323840045928955
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,4,2,power_law_1.01,1.289047088623047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,4,2,power_law_1.2,0.2060479927062988
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,4,2,power_law_1.2,0.30113792419433594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,4,2,power_law_1.01,1.7181759643554688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,4,2,power_law_1.2,0.387011833190918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,4,2,power_law_1.01,7.049019165039063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,4,2,power_law_1.2,0.5046195220947266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,4,2,power_law_1.2,0.530835189819336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,4,2,power_law_1.2,0.5082009506225587
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,4,2,power_law_1.2,0.5648217773437499
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,4,2,power_law_1.2,0.5792243194580078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,4,2,power_law_1.01,2.380431365966797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,4,2,power_law_1.2,0.8306444549560548
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,4,2,power_law_1.2,0.8613465881347656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,4,2,power_law_1.2,0.9547917175292969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,4,2,power_law_1.01,3.038005676269531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,4,2,power_law_1.2,0.9758297729492188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,4,2,power_law_1.2,1.0173670196533204
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,4,2,power_law_1.2,1.1570162963867188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,4,2,power_law_1.2,0.8305049896240234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,4,2,power_law_1.2,1.229100799560547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,4,2,power_law_1.2,1.3815245056152343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,4,2,power_law_1.2,1.5865049743652344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,4,2,power_law_1.2,0.05274496078491211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,4,2,power_law_1.2,0.07116543769836425
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,4,2,power_law_1.2,0.08767744064331054
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,4,2,power_law_1.2,0.13782400131225586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,4,2,power_law_1.2,2.347869415283203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,4,2,power_law_1.2,0.16382720947265625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,4,2,power_law_1.2,0.20801664352416993
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,4,2,power_law_1.2,0.27355392456054684
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,4,2,power_law_1.2,0.2950592041015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,4,2,power_law_1.2,1.8991795349121094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,4,2,power_law_1.2,0.30931327819824217
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,4,2,power_law_1.2,0.3155187225341797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,4,2,power_law_1.2,0.338152961730957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,4,2,power_law_1.2,0.3605964660644531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,4,2,power_law_1.2,0.3817830276489258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,4,2,power_law_1.2,0.39485183715820316
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,4,2,power_law_1.2,0.41477504730224607
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,4,2,power_law_1.2,3.0782183837890624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,4,2,power_law_1.2,4.1188916015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,4,2,power_law_1.2,0.42760448455810546
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,4,2,power_law_1.2,0.4424934387207031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,4,2,power_law_1.2,0.47731201171875004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,4,2,power_law_1.2,0.5401548767089844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,4,2,power_law_1.2,0.5977254486083984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,4,2,power_law_1.2,0.6953702545166015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,4,2,power_law_1.2,0.8301261138916015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,8,1,balanced,0.07284224033355713
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,4,2,power_law_1.2,1.0170329284667967
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,4,2,power_law_1.2,5.724580078124999
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,8,1,balanced,0.0859001636505127
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,8,1,balanced,0.11820351600646972
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,4,2,power_law_1.2,1.3487834167480468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,8,1,balanced,0.1909324836730957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,8,1,balanced,0.4186809539794922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,4,2,power_law_1.2,1.6616423034667966
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,8,1,balanced,0.7080403137207031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,8,1,balanced,0.7080486297607422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,8,1,balanced,0.7119487762451172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,4,2,power_law_1.2,2.450780181884766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,8,1,balanced,0.7193062591552735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,4,2,power_law_1.2,7.232323608398437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,8,1,balanced,0.7270233917236328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,8,1,balanced,0.7263993835449218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,4,2,power_law_1.2,3.066220703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,8,1,balanced,0.7325997161865234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,8,1,balanced,0.7421971130371093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,8,1,balanced,0.7139008331298828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,8,1,balanced,1.0317273712158204
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,8,1,balanced,1.0409996795654297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,8,1,balanced,1.0564083099365233
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,8,1,balanced,1.1105606079101562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,8,1,balanced,1.1652377319335936
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,8,1,balanced,1.2890777587890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,8,1,balanced,1.4129408264160157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,8,1,balanced,0.0493446397781372
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,8,1,balanced,0.06400256156921387
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,8,1,balanced,1.953280029296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,8,1,balanced,0.08522368431091308
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,8,1,balanced,1.9670988464355468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,8,1,balanced,0.13733759880065918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,8,1,balanced,0.24700672149658204
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,8,1,balanced,0.458331527709961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,8,1,balanced,0.4602297592163086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,8,1,balanced,0.4641574478149414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,8,1,balanced,2.660689392089844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,8,1,balanced,0.4649190521240234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,8,1,balanced,0.4686483383178711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,8,1,balanced,0.4743398284912109
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,8,1,balanced,0.48061569213867184
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,8,1,balanced,0.4833523178100586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,8,1,balanced,0.5680140686035157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,8,1,balanced,0.5749427032470703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,8,1,balanced,0.5828876876831055
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,8,1,balanced,0.594312973022461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,8,1,balanced,3.4087814331054687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,8,1,balanced,0.6121184158325195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,8,1,balanced,0.6575820922851563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,8,1,balanced,0.7010240173339843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,8,1,balanced,0.78761474609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,8,1,balanced,5.067091979980469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,8,1,balanced,0.48593791961669924
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,8,1,balanced,0.8871775817871093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,8,1,power_law_1.01,0.07422143936157226
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,8,1,power_law_1.01,0.08525504112243652
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,8,1,power_law_1.01,0.11668864250183106
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,8,1,balanced,1.3885279846191405
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,8,1,power_law_1.01,0.1898963165283203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,8,1,balanced,1.5940071105957032
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,8,1,power_law_1.01,0.29929983139038085
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,8,1,power_law_1.01,0.432856330871582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,8,1,power_law_1.01,0.5250118255615235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,8,1,power_law_1.01,0.5465599822998046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,8,1,power_law_1.01,0.5615283203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,8,1,power_law_1.01,0.5793619155883789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,8,1,balanced,2.3210111999511716
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,8,1,power_law_1.01,0.647171173095703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,8,1,power_law_1.01,0.667291488647461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,8,1,power_law_1.01,0.6707186889648438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,8,1,power_law_1.01,0.6886624145507813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,8,1,balanced,6.862867431640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,8,1,power_law_1.01,0.9141254425048828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,8,1,power_law_1.01,0.9320403289794921
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,8,1,power_law_1.01,0.99168701171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,8,1,power_law_1.01,1.0907193756103515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,8,1,power_law_1.01,1.1837702178955078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,8,1,balanced,2.9824761962890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,8,1,power_law_1.01,1.319943084716797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,8,1,power_law_1.01,1.4931872558593748
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,8,1,power_law_1.01,0.04936511993408203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,8,1,power_law_1.01,1.7976364135742187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,8,1,power_law_1.01,0.06345856189727783
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,8,1,power_law_1.01,0.08474880218505859
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,8,1,power_law_1.01,2.1572364807128905
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,8,1,power_law_1.01,0.13909055709838866
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,8,1,power_law_1.01,0.19985088348388672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,8,1,power_law_1.01,0.25471935272216795
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,8,1,power_law_1.01,0.3489459228515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,8,1,power_law_1.01,0.3631020736694336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,8,1,power_law_1.01,0.37165504455566406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,8,1,power_law_1.01,0.3821452713012695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,8,1,power_law_1.01,2.894328918457031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,8,1,power_law_1.01,0.4039449691772461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,8,1,power_law_1.01,0.41522304534912113
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,8,1,power_law_1.01,0.4200416183471679
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,8,1,power_law_1.01,0.4274508666992188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,8,1,power_law_1.01,0.5194348907470703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,8,1,power_law_1.01,3.6532888793945313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,8,1,power_law_1.01,0.5215795135498047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,8,1,power_law_1.01,0.537644157409668
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,8,1,power_law_1.01,0.5611846542358399
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,8,1,power_law_1.01,0.6036767959594727
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,8,1,power_law_1.01,0.6889798736572266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,8,1,power_law_1.01,0.75328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,8,1,power_law_1.01,0.9190048217773438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,8,1,power_law_1.01,1.0769407653808594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,8,1,power_law_1.2,0.07255680084228515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,8,1,power_law_1.01,5.210813598632813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,8,1,power_law_1.2,0.08557248115539551
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,8,1,power_law_1.2,0.11640447616577147
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,8,1,power_law_1.01,1.4222982788085938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,8,1,power_law_1.2,0.17868799209594727
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,8,1,power_law_1.2,0.2667334365844726
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,8,1,power_law_1.01,1.7830982971191407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,8,1,power_law_1.2,0.427852783203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,8,1,power_law_1.2,0.5170470428466797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,8,1,power_law_1.2,0.5390374374389648
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,8,1,power_law_1.2,0.551349105834961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,8,1,power_law_1.2,0.5920902252197265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,8,1,power_law_1.01,2.5012864685058593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,8,1,power_law_1.2,0.6541286468505859
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,8,1,power_law_1.2,0.6608525085449218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,8,1,power_law_1.2,0.6728230285644531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,8,1,power_law_1.2,0.7100371551513672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,8,1,power_law_1.01,6.718161010742188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,8,1,power_law_1.01,3.207198791503906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,8,1,power_law_1.2,0.9356652832031249
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,8,1,power_law_1.2,0.9614701080322267
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,8,1,power_law_1.2,1.0136390686035157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,8,1,power_law_1.2,1.154588165283203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,8,1,power_law_1.2,1.2146975708007812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,8,1,power_law_1.2,1.3707481384277345
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,8,1,power_law_1.2,1.515028533935547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,8,1,power_law_1.2,0.04976255893707275
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,8,1,power_law_1.2,1.8399999999999999
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,8,1,power_law_1.2,0.06347008228302002
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,8,1,power_law_1.2,0.08508992195129395
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,8,1,power_law_1.2,2.2108248901367187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,8,1,power_law_1.2,0.1306713581085205
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,8,1,power_law_1.2,0.18815040588378906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,8,1,power_law_1.2,0.24339008331298828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,8,1,power_law_1.2,0.3428806304931641
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,8,1,power_law_1.2,0.355849609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,8,1,power_law_1.2,0.36463489532470705
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,8,1,power_law_1.2,0.38603328704833983
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,8,1,power_law_1.2,2.953261413574219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,8,1,power_law_1.2,0.4006367874145508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,8,1,power_law_1.2,0.41086463928222655
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,8,1,power_law_1.2,0.4216915130615234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,8,1,power_law_1.2,0.43325950622558596
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,8,1,power_law_1.2,0.5255545425415039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,8,1,power_law_1.2,0.5335168075561524
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,8,1,power_law_1.2,0.5434982299804687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,8,1,power_law_1.2,3.716739196777344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,8,1,power_law_1.2,0.5901132965087891
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,8,1,power_law_1.2,0.6428166198730468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,8,1,power_law_1.2,0.7122937774658202
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,8,1,power_law_1.2,0.7924569702148437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,8,1,power_law_1.2,0.9495661163330078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,8,1,power_law_1.2,1.125266571044922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,1,8,balanced,0.06345727920532226
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,1,8,balanced,0.07253119945526124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,8,1,power_law_1.2,5.262927856445312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,8,1,power_law_1.2,1.4389254760742187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,1,8,balanced,0.15919232368469238
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,1,8,balanced,0.3570739364624024
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,1,8,balanced,0.6078220748901367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,8,1,power_law_1.2,1.8126559448242188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,1,8,balanced,0.09841919898986816
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,1,8,balanced,0.8995430755615235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,1,8,balanced,0.9049088287353516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,1,8,balanced,0.9111910247802735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,1,8,balanced,0.9142425537109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,8,1,power_law_1.2,6.917228393554687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,8,1,power_law_1.2,2.5466131591796874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,1,8,balanced,0.9051289367675782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,1,8,balanced,0.9137497711181641
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,1,8,balanced,0.9190067291259766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,1,8,balanced,0.9312857818603515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,1,8,balanced,0.951151351928711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,1,8,balanced,0.9683634948730468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,1,8,balanced,0.9859417724609376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,1,8,balanced,1.040953598022461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,1,8,balanced,1.0981401824951171
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,8,1,power_law_1.2,3.3506527709960934
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,1,8,balanced,1.1936000061035155
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,1,8,balanced,0.039443199634552
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,1,8,balanced,1.2744230651855468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,1,8,balanced,0.048081917762756346
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,1,8,balanced,0.06349311828613281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,1,8,balanced,1.5015461730957032
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,1,8,balanced,0.0981056022644043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,1,8,balanced,0.16958463668823243
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,1,8,balanced,1.7019699096679688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,1,8,balanced,0.31440000534057616
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,1,8,balanced,0.34097919464111326
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,1,8,balanced,0.3441958236694336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,1,8,balanced,0.3500172805786133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,1,8,balanced,0.3500006484985351
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,1,8,balanced,2.413721618652344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,1,8,balanced,0.3410508728027344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,1,8,balanced,0.34394622802734376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,1,8,balanced,0.3463833618164062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,1,8,balanced,0.35261566162109376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,1,8,balanced,0.3599308776855469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,1,8,balanced,3.0578842163085938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,1,8,balanced,0.36825473785400387
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,1,8,balanced,0.3884403228759766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,1,8,balanced,0.41782783508300786
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,1,8,balanced,0.4444659042358398
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,1,8,balanced,0.4937907028198242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,1,8,balanced,0.5478758239746094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,1,8,balanced,0.6604083251953126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,1,8,balanced,0.7913740539550782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,1,8,power_law_1.01,0.15590784072875977
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,1,8,balanced,4.584046020507812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,1,8,balanced,1.098440933227539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,1,8,power_law_1.01,0.3473062515258789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,1,8,power_law_1.01,0.4042982482910157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,1,8,balanced,1.3285772705078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,1,8,power_law_1.01,0.47809280395507814
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,1,8,power_law_1.01,0.5251200103759766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,1,8,power_law_1.01,0.5087411117553711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,1,8,balanced,1.9157530212402343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,1,8,power_law_1.01,0.7077913665771485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,1,8,balanced,5.943750610351563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,1,8,power_law_1.01,0.7734361267089843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,1,8,balanced,2.489637145996094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,1,8,power_law_1.01,0.7368678283691407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,1,8,power_law_1.01,0.7733708953857421
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,1,8,power_law_1.01,0.8081292724609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,1,8,power_law_1.01,0.8254924774169922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,1,8,power_law_1.01,0.8279987335205078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,1,8,power_law_1.01,0.8316620635986329
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,1,8,power_law_1.01,0.8913919830322266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,1,8,power_law_1.01,0.8926617431640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,1,8,power_law_1.01,0.9748377227783204
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,1,8,power_law_1.01,0.9829478454589843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,1,8,power_law_1.01,1.1120793914794922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,1,8,power_law_1.01,1.2682828521728515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,1,8,power_law_1.01,1.4279180908203126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,1,8,power_law_1.01,0.097194242477417
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,1,8,power_law_1.01,0.16746112823486328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,1,8,power_law_1.01,1.8081651306152344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,1,8,power_law_1.01,0.19486591339111328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,1,8,power_law_1.01,2.2912255859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,1,8,power_law_1.01,0.24962688446044923
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,1,8,power_law_1.01,0.2705817604064941
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,1,8,power_law_1.01,0.2632652854919434
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,1,8,power_law_1.01,0.2818355178833008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,1,8,power_law_1.01,0.2789708709716797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,1,8,power_law_1.01,0.28941568374633786
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,1,8,power_law_1.01,0.30449279785156247
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,1,8,power_law_1.01,0.31859455108642576
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,1,8,power_law_1.01,0.32077056884765626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,1,8,power_law_1.01,0.31775999069213867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,1,8,power_law_1.01,3.461996765136719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,1,8,power_law_1.01,0.32414081573486325
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,1,8,power_law_1.01,0.3313100814819336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,1,8,power_law_1.01,0.33424129486083987
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,1,8,power_law_1.01,0.3443174362182617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,1,8,power_law_1.01,0.380761604309082
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,1,8,power_law_1.01,4.174997863769532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,1,8,power_law_1.01,0.4969254302978515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,1,8,power_law_1.01,0.5652044677734376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,1,8,power_law_1.01,0.6906969451904297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,1,8,power_law_1.01,0.8714137268066405
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,1,8,power_law_1.01,0.41840385437011723
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,1,8,power_law_1.2,0.1582387161254883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,1,8,power_law_1.01,1.2315853118896485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,1,8,power_law_1.2,0.3459968185424805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,1,8,power_law_1.2,0.3544550323486328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,1,8,power_law_1.01,1.600861511230469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,1,8,power_law_1.2,0.44394496917724613
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,1,8,power_law_1.01,6.18630126953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,1,8,power_law_1.2,0.4938675308227539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,1,8,power_law_1.2,0.5505484771728516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,1,8,power_law_1.2,0.6698995208740234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,1,8,power_law_1.01,2.4599769592285154
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,1,8,power_law_1.2,0.7168870544433593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,1,8,power_law_1.2,0.7415116882324219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,1,8,power_law_1.2,0.7366067504882812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,1,8,power_law_1.2,0.7852198028564453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,1,8,power_law_1.01,2.9808666992187502
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,1,8,power_law_1.2,0.7969203186035156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,1,8,power_law_1.2,0.8492864227294922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,1,8,power_law_1.01,8.095072021484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,1,8,power_law_1.2,0.8542195129394532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,1,8,power_law_1.2,0.8917772674560547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,1,8,power_law_1.2,0.9229325103759767
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,1,8,power_law_1.2,0.949438705444336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,1,8,power_law_1.2,1.0662655639648437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,1,8,power_law_1.2,1.2165145874023438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,1,8,power_law_1.2,1.3303897094726562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,1,8,power_law_1.2,1.6311372375488282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,1,8,power_law_1.2,0.09665663719177246
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,1,8,power_law_1.2,0.16843776702880858
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,1,8,power_law_1.2,2.2413221740722653
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,1,8,power_law_1.2,2.4654847717285158
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,1,8,power_law_1.2,0.23390975952148438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,1,8,power_law_1.2,0.26129663467407227
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,1,8,power_law_1.2,0.2808268737792969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,1,8,power_law_1.2,0.26957824707031247
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,1,8,power_law_1.2,0.2707263946533203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,1,8,power_law_1.2,0.1633945655822754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,1,8,power_law_1.2,0.2887283134460449
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,1,8,power_law_1.2,0.3052249526977539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,1,8,power_law_1.2,0.3197913551330566
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,1,8,power_law_1.2,0.31390592575073245
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,1,8,power_law_1.2,0.3236454391479492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,1,8,power_law_1.2,0.3239295959472656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,1,8,power_law_1.2,0.33529727935791015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,1,8,power_law_1.2,0.3398720169067383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,1,8,power_law_1.2,0.34547584533691406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,1,8,power_law_1.2,0.40943744659423825
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,1,8,power_law_1.2,3.6929318237304685
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,1,8,power_law_1.2,0.4438860702514648
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,1,8,power_law_1.2,4.757507934570312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,1,8,power_law_1.2,0.5214361572265626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,1,8,power_law_1.2,0.5685196685791015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,1,8,power_law_1.2,0.7739225769042969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,1,8,power_law_1.2,0.9943641662597656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,2,4,balanced,0.05927680015563965
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,1,8,power_law_1.2,1.360809020996094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,2,4,balanced,0.07020415782928466
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,2,4,balanced,0.09691391944885254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,1,8,power_law_1.2,1.709180145263672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,2,4,balanced,0.15758463859558106
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,2,4,balanced,0.3509299087524414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,2,4,balanced,0.6039360046386719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,2,4,balanced,0.6050585556030273
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,1,8,power_law_1.2,7.7465625000000005
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,1,8,power_law_1.2,2.7094949340820316
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,2,4,balanced,0.8734502410888672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,2,4,balanced,0.8778214263916017
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,2,4,balanced,0.8824076843261718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,2,4,balanced,0.899609603881836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,2,4,balanced,0.8999820709228515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,2,4,balanced,0.6049280166625977
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,1,8,power_law_1.2,3.5983526611328123
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,2,4,balanced,0.9059801483154297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,2,4,balanced,0.9229618835449219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,2,4,balanced,0.9350962829589843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,2,4,balanced,0.9564403533935547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,2,4,balanced,1.0110975646972657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,2,4,balanced,1.0742591857910155
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,1,8,power_law_1.2,10.355202636718749
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,2,4,balanced,1.1712998199462892
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,2,4,balanced,1.256616973876953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,2,4,balanced,1.4897946166992189
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,2,4,balanced,0.0480076789855957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,2,4,balanced,0.06381440162658691
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,2,4,balanced,0.09759488105773925
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,2,4,balanced,0.16970752716064452
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,2,4,balanced,0.308984317779541
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,2,4,balanced,0.3126553535461426
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,2,4,balanced,0.03589119911193848
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,2,4,balanced,2.3545625305175784
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,2,4,balanced,0.31533056259155273
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,2,4,balanced,1.6682150268554685
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,2,4,balanced,0.33955200195312496
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,2,4,balanced,0.3404390335083008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,2,4,balanced,0.34576255798339844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,2,4,balanced,0.3444287872314453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,2,4,balanced,2.972588806152344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,2,4,balanced,0.3475596618652344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,2,4,balanced,0.35311744689941404
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,2,4,balanced,0.3579532623291016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,2,4,balanced,0.3648588943481445
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,2,4,balanced,0.3788224029541015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,2,4,balanced,0.4126707077026367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,2,4,balanced,0.42731647491455077
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,2,4,balanced,0.47636863708496097
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,2,4,balanced,0.5343231964111328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,2,4,balanced,0.6149657440185547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,2,4,balanced,0.7324671936035156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,2,4,power_law_1.01,0.09422592163085938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,2,4,balanced,4.392697448730469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,2,4,power_law_1.01,0.155861759185791
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,2,4,balanced,1.0460249328613282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,2,4,power_law_1.01,0.21869056701660156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,2,4,power_law_1.01,0.28065664291381837
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,2,4,balanced,1.245629425048828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,2,4,power_law_1.01,0.3724544143676758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,2,4,power_law_1.01,0.4048448181152343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,2,4,power_law_1.01,0.5354240036010742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,2,4,balanced,1.763504638671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,2,4,power_law_1.01,0.5553510284423828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,2,4,balanced,5.5340814208984375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,2,4,power_law_1.01,0.7296307373046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,2,4,power_law_1.01,0.764843521118164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,2,4,power_law_1.01,0.7577382659912109
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,2,4,balanced,2.339539184570312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,2,4,power_law_1.01,0.7867302703857422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,2,4,power_law_1.01,0.7841292572021484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,2,4,power_law_1.01,0.8270182037353516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,2,4,power_law_1.01,0.8658751678466796
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,2,4,power_law_1.01,0.8893529510498046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,2,4,power_law_1.01,0.9320499420166015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,2,4,power_law_1.01,1.0399667358398437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,2,4,power_law_1.01,1.0803929901123046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,2,4,power_law_1.01,1.2006578826904297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,2,4,power_law_1.01,1.3376972961425782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,2,4,power_law_1.01,0.06209663867950439
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,2,4,power_law_1.01,1.6537164306640626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,2,4,power_law_1.01,0.09641728401184083
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,2,4,power_law_1.01,0.12314496040344239
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,2,4,power_law_1.01,0.14723584175109863
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,2,4,power_law_1.01,2.0698150634765624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,2,4,power_law_1.01,0.19245824813842774
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,2,4,power_law_1.01,0.21534591674804687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,2,4,power_law_1.01,0.2749760055541992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,2,4,power_law_1.01,0.2932595252990723
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,2,4,power_law_1.01,0.2913599967956543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,2,4,power_law_1.01,0.29616895675659183
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,2,4,power_law_1.01,2.8276263427734376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,2,4,power_law_1.01,0.3011084747314453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,2,4,power_law_1.01,0.31276927947998046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,2,4,power_law_1.01,0.31270912170410153
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,2,4,power_law_1.01,0.3282086563110352
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,2,4,power_law_1.01,0.33871105194091794
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,2,4,power_law_1.01,0.33331199645996096
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,2,4,power_law_1.01,0.3478412628173828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,2,4,power_law_1.01,0.40540542602539065
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,2,4,power_law_1.01,3.861873779296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,2,4,power_law_1.01,0.4325030517578125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,2,4,power_law_1.01,0.4956825637817383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,2,4,power_law_1.01,0.5626943969726563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,2,4,power_law_1.01,0.6819596862792968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,2,4,power_law_1.01,0.8353958129882812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,2,4,power_law_1.2,0.09473919868469238
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,2,4,power_law_1.01,1.1423705291748045
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,2,4,power_law_1.2,0.1551743984222412
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,2,4,power_law_1.01,5.546763305664062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,2,4,power_law_1.2,0.19029504776000977
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,2,4,power_law_1.01,1.4404428100585938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,2,4,power_law_1.2,0.31281280517578125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,2,4,power_law_1.2,0.35925758361816407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,2,4,power_law_1.2,0.4286540985107422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,2,4,power_law_1.2,0.5337369537353516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,2,4,power_law_1.2,0.531297264099121
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,2,4,power_law_1.01,2.146730194091797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,2,4,power_law_1.2,0.7346969604492187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,2,4,power_law_1.2,0.747119369506836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,2,4,power_law_1.01,6.912415771484374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,2,4,power_law_1.2,0.7843571472167968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,2,4,power_law_1.2,0.798229751586914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,2,4,power_law_1.2,0.8035225677490235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,2,4,power_law_1.01,2.71825927734375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,2,4,power_law_1.2,0.8451238250732421
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,2,4,power_law_1.2,0.866983642578125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,2,4,power_law_1.2,0.8736153411865235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,2,4,power_law_1.2,0.9818214416503906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,2,4,power_law_1.2,1.0235814666748047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,2,4,power_law_1.2,1.0822054290771486
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,2,4,power_law_1.2,1.2457266998291014
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,2,4,power_law_1.2,1.4469786071777344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,2,4,power_law_1.2,0.060921602249145515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,2,4,power_law_1.2,1.8054566955566407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,2,4,power_law_1.2,0.09628543853759766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,2,4,power_law_1.2,0.11041152000427246
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,2,4,power_law_1.2,0.1430399990081787
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,2,4,power_law_1.2,0.17534080505371094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,2,4,power_law_1.2,0.2116787147521973
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,2,4,power_law_1.2,2.2872653198242188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,2,4,power_law_1.2,0.2843699264526367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,2,4,power_law_1.2,0.27725183486938476
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,2,4,power_law_1.2,0.2729254341125488
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,2,4,power_law_1.2,0.30712831497192383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,2,4,power_law_1.2,3.1434854125976566
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,2,4,power_law_1.2,0.30254848480224605
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,2,4,power_law_1.2,0.3127359962463379
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,2,4,power_law_1.2,0.32075775146484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,2,4,power_law_1.2,0.32889984130859373
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,2,4,power_law_1.2,0.3355007934570312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,2,4,power_law_1.2,0.3378700637817383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,2,4,power_law_1.2,0.35884033203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,2,4,power_law_1.2,0.41187072753906245
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,2,4,power_law_1.2,3.7918426513671877
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,2,4,power_law_1.2,0.4479359817504883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,2,4,power_law_1.2,0.5074252700805664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,2,4,power_law_1.2,0.5822553634643555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,2,4,power_law_1.2,0.7180403137207032
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,4,2,balanced,0.05803008079528808
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,4,2,balanced,0.0678937578201294
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,2,4,power_law_1.2,1.233406753540039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,2,4,power_law_1.2,5.216029663085938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,4,2,balanced,0.09528063774108887
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,4,2,balanced,0.16070016860961914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,2,4,power_law_1.2,1.650872344970703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,4,2,balanced,0.35156864166259766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,4,2,balanced,0.598917121887207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,2,4,power_law_1.2,0.9272383880615234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,4,2,balanced,0.600134391784668
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,4,2,balanced,0.6079667282104493
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,4,2,balanced,0.6066739273071289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,2,4,power_law_1.2,2.2113037109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,4,2,balanced,0.6070272064208984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,4,2,balanced,0.6144435119628906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,4,2,balanced,0.8723686218261719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,4,2,balanced,0.8788134765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,2,4,power_law_1.2,3.000232849121094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,4,2,balanced,0.8925337219238282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,4,2,balanced,0.909991683959961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,4,2,balanced,0.9218470764160156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,4,2,balanced,0.9393484497070312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,2,4,power_law_1.2,7.828816528320313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,4,2,balanced,0.9855244445800782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,4,2,balanced,1.038620147705078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,4,2,balanced,1.1399974060058593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,4,2,balanced,1.2304883575439454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,4,2,balanced,0.0331276798248291
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,4,2,balanced,0.04769152164459228
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,4,2,balanced,1.5146815490722656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,4,2,balanced,0.06345471858978272
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,4,2,balanced,0.10141568183898926
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,4,2,balanced,1.7151641845703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,4,2,balanced,0.1811814308166504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,4,2,balanced,0.3286707305908203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,4,2,balanced,0.33176830291748044
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,4,2,balanced,0.3327769470214844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,4,2,balanced,2.3351539611816405
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,4,2,balanced,0.33663871765136716
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,4,2,balanced,0.34007553100585936
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,4,2,balanced,0.34326145172119144
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,4,2,balanced,0.38621440887451175
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,4,2,balanced,0.38546047210693357
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,4,2,balanced,3.157196655273437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,4,2,balanced,0.38916606903076173
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,4,2,balanced,0.3933951950073242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,4,2,balanced,0.3989683151245117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,4,2,balanced,0.40900096893310545
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,4,2,balanced,0.4464563369750977
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,4,2,balanced,0.45254142761230465
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,4,2,balanced,0.4907583999633789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,4,2,balanced,0.539153938293457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,4,2,balanced,0.6276147079467773
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,4,2,balanced,0.7244902038574219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,4,2,balanced,4.349847106933593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,4,2,power_law_1.01,0.06778111934661865
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,4,2,balanced,1.0693414306640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,4,2,power_law_1.01,0.094717435836792
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,4,2,power_law_1.01,0.13117695808410645
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,4,2,power_law_1.01,0.18688640594482422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,4,2,balanced,1.2477286529541014
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,4,2,power_law_1.01,0.3178303909301758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,4,2,power_law_1.01,0.3641459274291992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,4,2,power_law_1.01,0.4578022384643555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,4,2,balanced,5.509926147460938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,4,2,power_law_1.01,0.499818229675293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,4,2,balanced,1.8033651733398437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,4,2,power_law_1.01,0.4959270477294922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,4,2,power_law_1.01,0.5004684829711914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,4,2,power_law_1.01,0.5299827194213866
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,4,2,power_law_1.01,0.7354879760742188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,4,2,balanced,2.381378631591797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,4,2,power_law_1.01,0.8046527862548828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,4,2,power_law_1.01,0.7872089385986328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,4,2,power_law_1.01,0.8282061004638672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,4,2,power_law_1.01,0.8448973083496094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,4,2,power_law_1.01,0.8736473846435547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,4,2,power_law_1.01,0.9916313934326173
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,4,2,power_law_1.01,1.03159423828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,4,2,power_law_1.01,1.1461004638671874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,4,2,power_law_1.01,1.2769920349121093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,4,2,power_law_1.01,0.04761600017547608
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,4,2,power_law_1.01,1.5559487915039063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,4,2,power_law_1.01,0.06278656005859376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,4,2,power_law_1.01,0.0816806411743164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,4,2,power_law_1.01,1.890476837158203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,4,2,power_law_1.01,0.12900735855102538
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,4,2,power_law_1.01,0.16119808197021485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,4,2,power_law_1.01,0.19961984634399416
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,4,2,power_law_1.01,0.26965120315551755
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,4,2,power_law_1.01,0.2616665649414062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,4,2,power_law_1.01,2.581268615722656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,4,2,power_law_1.01,0.2774784088134766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,4,2,power_law_1.01,0.2824140739440918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,4,2,power_law_1.01,0.2945574378967285
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,4,2,power_law_1.01,0.3359731292724609
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,4,2,power_law_1.01,0.3566681671142578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,4,2,power_law_1.01,0.357498893737793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,4,2,power_law_1.01,3.2451480102539065
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,4,2,power_law_1.01,0.3684415817260742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,4,2,power_law_1.01,0.36559486389160156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,4,2,power_law_1.01,0.37634304046630856
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,4,2,power_law_1.01,0.40610431671142583
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,4,2,power_law_1.01,0.45273983001708984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,4,2,power_law_1.01,0.5050201416015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,4,2,power_law_1.01,0.57731201171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,4,2,power_law_1.01,0.7010880279541015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,4,2,power_law_1.01,0.8756403350830079
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,4,2,power_law_1.01,4.5124813842773435
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,4,2,power_law_1.2,0.06792704105377198
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,4,2,power_law_1.2,0.0941427230834961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,4,2,power_law_1.01,1.1565106964111327
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,4,2,power_law_1.2,0.17939071655273436
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,4,2,power_law_1.01,1.457222442626953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,4,2,power_law_1.2,0.27819648742675784
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,4,2,power_law_1.2,0.343823356628418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,4,2,power_law_1.2,0.45759231567382813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,4,2,power_law_1.2,0.4544819259643555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,4,2,power_law_1.2,0.48378623962402345
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,4,2,power_law_1.2,0.12129407882690431
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,4,2,power_law_1.2,0.49916545867919926
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,4,2,power_law_1.01,5.940568237304687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,4,2,power_law_1.2,0.5457254409790039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,4,2,power_law_1.01,2.6944116210937503
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,4,2,power_law_1.2,0.7507814025878906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,4,2,power_law_1.2,0.7839564514160157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,4,2,power_law_1.2,0.8182911682128907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,4,2,power_law_1.2,0.8431206512451173
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,4,2,power_law_1.01,2.0032537841796874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,4,2,power_law_1.2,0.8753600311279296
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,4,2,power_law_1.2,0.9261055755615235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,4,2,power_law_1.2,1.0410009765625001
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,4,2,power_law_1.2,1.191088638305664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,4,2,power_law_1.2,1.3320307922363281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,4,2,power_law_1.2,1.6360000610351562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,4,2,power_law_1.2,0.05129856109619141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,4,2,power_law_1.2,1.0043033599853515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,4,2,power_law_1.2,2.02830078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,4,2,power_law_1.2,0.06425600051879883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,4,2,power_law_1.2,0.08033151626586914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,4,2,power_law_1.2,0.12684160232543945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,4,2,power_law_1.2,2.7368218994140627
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,4,2,power_law_1.2,0.14919039726257324
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,4,2,power_law_1.2,0.18185216903686524
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,4,2,power_law_1.2,0.2694425582885742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,4,2,power_law_1.2,0.26494592666625977
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,4,2,power_law_1.2,3.3192807006835934
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,4,2,power_law_1.2,0.2731161689758301
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,4,2,power_law_1.2,0.28820863723754886
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,4,2,power_law_1.2,0.29287424087524416
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,4,2,power_law_1.2,0.3408371353149414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,4,2,power_law_1.2,0.35728897094726564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,4,2,power_law_1.2,0.36184318542480465
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,4,2,power_law_1.2,0.3695142364501953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,4,2,power_law_1.2,0.3887321472167969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,4,2,power_law_1.2,0.4160128021240235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,4,2,power_law_1.2,0.47646720886230465
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,4,2,power_law_1.2,0.5296550369262695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,4,2,power_law_1.2,0.6096883010864258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,4,2,power_law_1.2,0.7382592010498047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,4,2,power_law_1.2,0.36874496459960937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,4,2,power_law_1.2,4.530922241210938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,8,1,balanced,0.06415103912353516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,4,2,power_law_1.2,0.885902099609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,8,1,balanced,0.07791232109069825
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,8,1,balanced,0.10439295768737793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,8,1,balanced,0.16466943740844725
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,4,2,power_law_1.2,1.1947814178466796
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,8,1,balanced,0.3590982437133789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,4,2,power_law_1.2,1.5113395690917968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,8,1,balanced,0.6134924697875976
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,8,1,balanced,0.6118412780761718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,8,1,balanced,0.6152883148193359
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,4,2,power_law_1.2,6.05304443359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,4,2,power_law_1.2,2.134122314453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,8,1,balanced,0.6182086563110352
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,8,1,balanced,0.6216704177856445
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,8,1,balanced,0.6272262573242188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,8,1,balanced,0.6350028610229492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,8,1,balanced,0.6394803237915039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,4,2,power_law_1.2,2.739427795410156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,8,1,balanced,0.6495359802246095
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,8,1,balanced,0.9038566589355469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,8,1,balanced,0.8976850891113282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,8,1,balanced,0.921506576538086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,8,1,balanced,0.9616569519042969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,8,1,balanced,1.0073503875732421
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,8,1,balanced,1.106195831298828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,8,1,balanced,1.2091487884521483
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,8,1,balanced,0.045758719444274905
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,8,1,balanced,0.05734720230102539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,8,1,balanced,0.07748991966247558
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,8,1,balanced,1.7216717529296877
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,8,1,balanced,0.12228480339050293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,8,1,balanced,1.652939453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,8,1,balanced,0.4193824005126953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,8,1,balanced,0.4007487869262695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,8,1,balanced,0.4180742263793945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,8,1,balanced,2.2940652465820315
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,8,1,balanced,0.40759552001953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,8,1,balanced,0.4057984161376953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,8,1,balanced,0.2158483123779297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,8,1,balanced,0.4114291381835938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,8,1,balanced,0.4167878341674805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,8,1,balanced,2.9118054199218752
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,8,1,balanced,0.4161286544799805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,8,1,balanced,0.4205420684814453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,8,1,balanced,0.49432384490966796
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,8,1,balanced,0.5004761505126953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,8,1,balanced,0.5077145767211915
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,8,1,balanced,0.5254963302612305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,8,1,balanced,0.5361862564086914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,8,1,balanced,0.5710015869140624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,8,1,balanced,0.6068076705932617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,8,1,balanced,0.6904275512695313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,8,1,balanced,0.7768128204345703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,8,1,power_law_1.01,0.06389952182769776
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,8,1,balanced,4.32163330078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,8,1,power_law_1.01,0.07709248065948486
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,8,1,balanced,1.200831985473633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,8,1,power_law_1.01,0.10322560310363768
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,8,1,power_law_1.01,0.1670047950744629
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,8,1,balanced,1.38841796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,8,1,power_law_1.01,0.2567033576965332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,8,1,power_law_1.01,0.37028289794921876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,8,1,power_law_1.01,0.4494617462158203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,8,1,power_law_1.01,0.46864192962646484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,8,1,balanced,1.97635009765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,8,1,power_law_1.01,0.481596794128418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,8,1,balanced,5.595972290039063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,8,1,power_law_1.01,0.5524403381347656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,8,1,power_law_1.01,0.5724396896362305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,8,1,power_law_1.01,0.5787705612182618
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,8,1,balanced,2.5800274658203124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,8,1,power_law_1.01,0.7974444580078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,8,1,power_law_1.01,0.4979673767089844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,8,1,power_law_1.01,0.8159814453124999
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,8,1,power_law_1.01,0.8608959960937501
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,8,1,power_law_1.01,0.5936883163452149
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,8,1,power_law_1.01,1.0199660491943359
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,8,1,power_law_1.01,1.1241529846191407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,8,1,power_law_1.01,1.2689311981201172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,8,1,power_law_1.01,0.9414246368408203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,8,1,power_law_1.01,1.5431878662109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,8,1,power_law_1.01,0.04539584159851075
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,8,1,power_law_1.01,0.0591321611404419
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,8,1,power_law_1.01,0.07704512119293214
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,8,1,power_law_1.01,1.8069056701660156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,8,1,power_law_1.01,0.12332415580749512
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,8,1,power_law_1.01,0.17814464569091798
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,8,1,power_law_1.01,0.22254400253295897
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,8,1,power_law_1.01,0.30282560348510745
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,8,1,power_law_1.01,0.3148819160461426
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,8,1,power_law_1.01,2.479754180908203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,8,1,power_law_1.01,0.3249631881713867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,8,1,power_law_1.01,0.332355842590332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,8,1,power_law_1.01,0.35050304412841793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,8,1,power_law_1.01,0.35982208251953124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,8,1,power_law_1.01,0.36616958618164064
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,8,1,power_law_1.01,0.3714931106567383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,8,1,power_law_1.01,0.4485177612304687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,8,1,power_law_1.01,3.105207824707031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,8,1,power_law_1.01,0.45402496337890624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,8,1,power_law_1.01,0.46675582885742184
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,8,1,power_law_1.01,0.49063743591308595
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,8,1,power_law_1.01,0.5283430480957032
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,8,1,power_law_1.01,0.5996076965332031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,8,1,power_law_1.01,0.6579475402832031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,8,1,power_law_1.01,0.8014470672607421
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,8,1,power_law_1.01,0.9490496063232422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,8,1,power_law_1.2,0.06513984203338623
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,8,1,power_law_1.01,4.4428103637695315
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,8,1,power_law_1.2,0.07729472160339355
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,8,1,power_law_1.2,0.10376383781433104
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,8,1,power_law_1.01,1.2358515167236328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,8,1,power_law_1.2,0.15669119834899903
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,8,1,power_law_1.2,0.22931327819824218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,8,1,power_law_1.01,1.552342987060547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,8,1,power_law_1.2,0.364851188659668
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,8,1,power_law_1.2,0.44275775909423826
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,8,1,power_law_1.2,0.46260608673095704
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,8,1,power_law_1.2,0.4710111999511719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,8,1,power_law_1.2,0.5079135894775391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,8,1,power_law_1.01,5.7357275390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,8,1,power_law_1.01,2.194615631103516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,8,1,power_law_1.2,0.5586406326293945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,8,1,power_law_1.2,0.5683302307128907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,8,1,power_law_1.2,0.5821088027954102
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,8,1,power_law_1.2,0.6067808151245118
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,8,1,power_law_1.01,2.795302429199219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,8,1,power_law_1.2,0.8139692687988281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,8,1,power_law_1.2,0.8409638214111329
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,8,1,power_law_1.2,0.8804742431640624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,8,1,power_law_1.2,0.9976396942138672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,8,1,power_law_1.2,1.0421356964111328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,8,1,power_law_1.2,1.1714304351806641
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,8,1,power_law_1.2,1.284267578125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,8,1,power_law_1.2,0.04464384078979492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,8,1,power_law_1.2,0.05792064189910888
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,8,1,power_law_1.2,1.5549356079101562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,8,1,power_law_1.2,0.07713215827941895
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,8,1,power_law_1.2,1.862410888671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,8,1,power_law_1.2,0.11678079605102538
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,8,1,power_law_1.2,0.16791488647460937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,8,1,power_law_1.2,0.2125632095336914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,8,1,power_law_1.2,0.2986432075500488
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,8,1,power_law_1.2,0.3116281509399414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,8,1,power_law_1.2,0.318338565826416
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,8,1,power_law_1.2,0.3369196701049805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,8,1,power_law_1.2,2.5775552368164063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,8,1,power_law_1.2,0.3481068801879883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,8,1,power_law_1.2,0.3578886413574219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,8,1,power_law_1.2,0.36555454254150394
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,8,1,power_law_1.2,0.3768966293334961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,8,1,power_law_1.2,3.1671060180664066
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,8,1,power_law_1.2,0.4602489471435547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,8,1,power_law_1.2,0.46459327697753905
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,8,1,power_law_1.2,0.47472576141357425
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,8,1,power_law_1.2,0.5179961776733398
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,8,1,power_law_1.2,0.5589497756958008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,8,1,power_law_1.2,0.6261663818359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,8,1,power_law_1.2,0.6893663787841797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,8,1,power_law_1.2,4.429912414550781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,8,1,power_law_1.2,0.867465591430664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,1,8,balanced,0.019330559968948363
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,1,8,balanced,0.020961279869079592
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,1,8,balanced,0.027488000392913818
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,1,8,balanced,0.039290881156921385
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,1,8,balanced,0.05367936134338379
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,1,8,balanced,0.07153408050537109
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,1,8,balanced,0.07505919933319091
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,1,8,balanced,0.07662591934204102
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,8,1,power_law_1.2,1.2524614715576172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,1,8,balanced,0.08080256462097168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,1,8,balanced,0.07309823989868164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,8,1,power_law_1.2,1.5757913208007812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,8,1,power_law_1.2,0.9766118621826172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,1,8,balanced,0.07811583995819091
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,1,8,balanced,0.08011263847351074
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,1,8,balanced,0.0847871971130371
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,8,1,power_law_1.2,5.753434448242187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,1,8,balanced,0.08685312271118165
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,1,8,balanced,0.0921292781829834
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,1,8,balanced,0.07926015853881836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,1,8,balanced,0.11699071884155274
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,1,8,balanced,0.1558668804168701
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,8,1,power_law_1.2,2.2146394348144534
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,1,8,balanced,0.1836761665344238
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,1,8,balanced,0.2577343940734863
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,1,8,balanced,0.3151462364196777
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,1,8,balanced,0.07580927848815919
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,1,8,balanced,0.02172800064086914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,1,8,balanced,0.5783014297485352
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,1,8,balanced,0.02188159942626953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,1,8,balanced,0.022602241039276123
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,8,1,power_law_1.2,2.850108947753906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,1,8,balanced,0.026616320610046384
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,1,8,balanced,0.04078464031219482
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,1,8,balanced,0.04897280216217041
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,1,8,balanced,0.05224063873291016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,1,8,balanced,0.05476352214813233
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,1,8,balanced,1.1239270019531251
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,1,8,balanced,0.05465856075286866
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,1,8,balanced,0.050343680381774905
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,1,8,balanced,0.05208703994750976
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,1,8,balanced,0.053726720809936526
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,1,8,balanced,0.058720002174377436
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,1,8,balanced,0.4565631866455078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,1,8,balanced,0.0650764799118042
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,1,8,balanced,0.07912191867828369
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,1,8,balanced,0.08922368049621582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,1,8,balanced,0.11544063568115234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,1,8,balanced,0.1112012767791748
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,1,8,balanced,0.19545984268188477
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,1,8,balanced,0.04771455764770508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,1,8,balanced,0.24023296356201174
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,1,8,balanced,0.3432588958740234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,1,8,balanced,0.4355712127685547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,1,8,balanced,0.05660927772521973
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,1,8,power_law_1.01,0.036894719600677486
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,1,8,balanced,0.060834560394287106
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,1,8,balanced,0.6466214752197266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,1,8,power_law_1.01,0.04065408229827881
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,1,8,power_law_1.01,0.04147071838378906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,1,8,power_law_1.01,0.04864384174346924
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,1,8,power_law_1.01,0.04592895984649658
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,1,8,balanced,0.8388057708740234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,1,8,power_law_1.01,0.07136256217956542
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,1,8,balanced,0.13928319931030272
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,1,8,power_law_1.01,0.07613440036773682
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,1,8,power_law_1.01,0.07678719997406006
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,1,8,power_law_1.01,0.07876607894897461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,1,8,power_law_1.01,0.07634175777435302
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,1,8,power_law_1.01,0.07700096130371094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,1,8,power_law_1.01,0.0834175968170166
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,1,8,power_law_1.01,0.08577280044555664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,1,8,power_law_1.01,0.09717120170593262
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,1,8,power_law_1.01,0.12500864028930664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,1,8,power_law_1.01,0.14505472183227539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,1,8,power_law_1.01,0.17370880126953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,1,8,power_law_1.01,0.23506303787231447
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,1,8,power_law_1.01,0.2990937614440918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,1,8,power_law_1.01,0.37083518981933594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,1,8,power_law_1.01,0.06817152023315429
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,1,8,power_law_1.01,0.08300543785095214
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,1,8,power_law_1.01,0.5643379211425781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,1,8,power_law_1.01,0.02265984058380127
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,1,8,power_law_1.01,0.02448256015777588
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,1,8,power_law_1.01,0.033712639808654785
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,1,8,power_law_1.01,0.7979955291748047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,1,8,power_law_1.01,0.035539200305938726
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,1,8,power_law_1.01,0.04397183895111084
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,1,8,power_law_1.01,0.04769663810729981
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,1,8,power_law_1.01,1.111568603515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,1,8,power_law_1.01,0.048958721160888674
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,1,8,power_law_1.01,0.05213696002960205
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,1,8,power_law_1.01,0.055020799636840825
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,1,8,power_law_1.01,0.07424511909484863
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,1,8,balanced,0.8591027069091798
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,1,8,power_law_1.01,0.050690560340881353
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,1,8,power_law_1.01,0.05307519912719727
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,1,8,power_law_1.01,0.05489664077758789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,1,8,power_law_1.01,0.05733119964599609
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,1,8,power_law_1.01,0.06342271804809571
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,1,8,power_law_1.01,0.05966207981109619
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,1,8,power_law_1.01,1.5435455322265625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,1,8,power_law_1.01,0.07806591987609864
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,1,8,power_law_1.01,0.09207296371459961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,1,8,power_law_1.01,0.11395199775695801
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,1,8,power_law_1.01,0.037735679149627686
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,1,8,power_law_1.01,0.18010368347167968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,1,8,power_law_1.01,0.23378431320190432
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,1,8,power_law_1.01,0.371569938659668
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,1,8,power_law_1.2,0.036729600429534916
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,1,8,power_law_1.01,0.46337024688720707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,1,8,power_law_1.2,0.037285120487213136
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,1,8,power_law_1.2,0.040017919540405275
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,1,8,power_law_1.01,0.055622401237487795
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,1,8,power_law_1.2,0.04693247795104981
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,1,8,power_law_1.01,0.6912870025634765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,1,8,power_law_1.2,0.042282238006591796
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,1,8,power_law_1.2,0.06871679782867432
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,1,8,power_law_1.2,0.07225599765777588
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,1,8,power_law_1.2,0.07454463958740234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,1,8,power_law_1.2,0.07689727783203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,1,8,power_law_1.01,0.9145830535888673
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,1,8,power_law_1.2,0.07774975776672363
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,1,8,power_law_1.2,0.08157312393188476
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,1,8,power_law_1.2,0.07679743766784668
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,1,8,power_law_1.2,0.07622144222259522
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,1,8,power_law_1.2,0.07986303806304931
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,1,8,power_law_1.2,0.08907391548156737
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,1,8,power_law_1.2,0.08571264266967774
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,1,8,power_law_1.2,0.09988736152648926
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,1,8,power_law_1.2,0.1306214427947998
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,1,8,power_law_1.2,0.15398271560668947
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,1,8,power_law_1.2,0.19599103927612305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,1,8,power_law_1.2,0.2572134399414062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,1,8,power_law_1.2,0.34345855712890627
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,1,8,power_law_1.2,0.02386944055557251
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,1,8,power_law_1.2,0.026463999748229983
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,1,8,power_law_1.2,0.44923774719238285
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,1,8,power_law_1.01,0.14245375633239746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,1,8,power_law_1.2,0.031439359188079836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,1,8,power_law_1.2,0.03585024118423462
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,1,8,power_law_1.2,0.034469120502471924
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,1,8,power_law_1.2,0.04434559822082519
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,1,8,power_law_1.2,0.048044800758361816
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,1,8,power_law_1.2,0.050472960472106934
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,1,8,power_law_1.2,0.05267583847045898
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,1,8,power_law_1.2,0.05437183856964112
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,1,8,power_law_1.2,0.06004992008209229
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,1,8,power_law_1.2,0.052203521728515626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,1,8,power_law_1.2,0.05570943832397461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,1,8,power_law_1.2,0.055349760055542
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,1,8,power_law_1.2,0.05945856094360351
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,1,8,power_law_1.2,0.06138112068176269
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,1,8,power_law_1.2,0.06681471824645996
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,1,8,power_law_1.2,0.08313216209411621
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,1,8,power_law_1.2,0.09678463935852051
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,1,8,power_law_1.2,0.12505215644836426
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,1,8,power_law_1.2,0.15492351531982423
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,1,8,power_law_1.2,0.6283276748657227
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,1,8,power_law_1.2,0.2207551956176758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,1,8,power_law_1.2,0.2574259185791016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,2,4,balanced,0.018705919981002808
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,2,4,balanced,0.01958400011062622
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,2,4,balanced,0.028161280155181885
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,2,4,balanced,0.038122239112854006
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,2,4,balanced,0.054630398750305176
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,2,4,balanced,0.056837120056152345
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,2,4,balanced,0.07344639778137207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,2,4,balanced,0.07607935905456543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,2,4,balanced,0.07762176036834717
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,1,8,power_law_1.2,0.9121305847167969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,2,4,balanced,0.08005760192871095
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,2,4,balanced,0.07214975833892823
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,2,4,balanced,0.0735103988647461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,2,4,balanced,0.07505919933319091
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,2,4,balanced,0.07658239841461181
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,2,4,balanced,0.0799616003036499
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,2,4,balanced,0.08249983787536622
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,2,4,balanced,0.08740736007690429
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,2,4,balanced,0.09925375938415527
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,2,4,balanced,0.11041024208068848
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,2,4,balanced,0.14534144401550292
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,2,4,balanced,0.16962560653686523
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,2,4,balanced,0.2422220802307129
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,2,4,balanced,0.29707519531250004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,2,4,balanced,0.43026817321777344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,1,8,power_law_1.2,0.4930534362792969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,2,4,balanced,0.5489139175415039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,2,4,balanced,0.021584639549255373
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,2,4,balanced,0.021945600509643552
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,2,4,balanced,0.8091072082519531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,2,4,balanced,0.023171839714050294
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,2,4,balanced,0.02671871900558472
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,1,8,power_law_1.2,0.39920513153076176
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,2,4,balanced,0.04382719993591309
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,2,4,balanced,1.058718719482422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,2,4,balanced,0.046577920913696294
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,2,4,balanced,0.05391615867614746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,2,4,balanced,0.05610367774963378
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,2,4,balanced,0.05785344123840332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,2,4,balanced,0.05947391986846924
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,2,4,balanced,0.05057024002075196
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,2,4,balanced,0.0516428804397583
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,2,4,balanced,0.053186559677124025
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,2,4,balanced,0.05619071960449219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,2,4,balanced,0.05952383995056152
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,2,4,balanced,0.06214655876159668
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,2,4,balanced,0.06549248218536377
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,2,4,balanced,0.07552768230438232
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,2,4,balanced,0.08497152328491211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,2,4,balanced,0.10598527908325195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,2,4,balanced,0.12530816078186036
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,2,4,balanced,0.18263168334960939
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,2,4,balanced,0.2202508735656738
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,2,4,balanced,0.31670656204223635
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,2,4,balanced,0.40154495239257815
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,1,8,power_law_1.2,1.35949951171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,2,4,power_law_1.01,0.023333120346069335
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,2,4,balanced,0.5942041778564453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,2,4,power_law_1.01,0.028387839794158938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,2,4,power_law_1.01,0.03612416028976441
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,2,4,power_law_1.01,0.04143104076385498
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,2,4,power_law_1.01,0.04017280101776123
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,2,4,balanced,0.7721894073486328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,2,4,power_law_1.01,0.05271935939788819
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,2,4,power_law_1.01,0.06981247901916504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,2,4,power_law_1.01,0.07256832122802734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,2,4,power_law_1.01,0.07585408210754394
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,2,4,power_law_1.01,0.0776204776763916
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,2,4,power_law_1.01,0.07953919887542724
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,2,4,power_law_1.01,0.07289216041564942
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,2,4,power_law_1.01,0.07493760108947753
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,2,4,power_law_1.01,0.07876607894897461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,2,4,power_law_1.01,0.08303359985351563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,2,4,power_law_1.01,0.0855948829650879
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,2,4,power_law_1.01,0.08994175910949706
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,2,4,power_law_1.01,0.10644736289978027
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,2,4,power_law_1.01,0.12491519927978516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,2,4,power_law_1.01,0.1648793601989746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,2,4,power_law_1.01,0.20742528915405273
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,2,4,power_law_1.01,0.2659161567687988
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,2,4,power_law_1.01,0.35640190124511717
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,2,4,power_law_1.01,0.49149185180664057
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,2,4,power_law_1.01,0.021914880275726318
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,2,4,power_law_1.01,0.6574092864990234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,2,4,power_law_1.01,0.025544960498809815
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,2,4,power_law_1.01,0.02897536039352417
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,2,4,power_law_1.01,0.032186880111694335
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,2,4,power_law_1.01,0.9306995391845703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,2,4,power_law_1.01,0.038416640758514406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,2,4,power_law_1.01,0.04505983829498291
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,2,4,power_law_1.01,0.05052671909332276
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,2,4,power_law_1.01,0.05371391773223877
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,2,4,power_law_1.01,0.05611135959625244
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,1,8,power_law_1.2,0.7776332855224609
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,2,4,power_law_1.01,1.2331852722167969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,2,4,power_law_1.01,0.056915202140808106
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,2,4,power_law_1.01,0.05076608180999755
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,2,4,power_law_1.01,0.05203455924987793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,2,4,power_law_1.01,0.05473152160644531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,2,4,power_law_1.01,0.0582144021987915
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,2,4,power_law_1.01,0.05955711841583252
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,2,4,power_law_1.01,0.06561535835266114
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,2,4,power_law_1.01,0.07723648071289062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,2,4,power_law_1.01,0.09093376159667968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,2,4,power_law_1.01,0.11887104034423827
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,2,4,power_law_1.01,0.14242815971374512
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,2,4,power_law_1.01,0.19027839660644533
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,2,4,power_law_1.01,0.23891199111938474
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,2,4,power_law_1.01,0.33611263275146486
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,2,4,power_law_1.2,0.020980479717254637
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,2,4,power_law_1.01,0.43825664520263674
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,2,4,power_law_1.01,0.05827072143554688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,2,4,power_law_1.2,0.030237441062927244
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,2,4,power_law_1.2,0.033625600337982176
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,2,4,power_law_1.01,0.6474380493164062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,2,4,power_law_1.2,0.04612607955932617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,2,4,power_law_1.2,0.05370240211486817
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,2,4,power_law_1.2,0.0706496000289917
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,2,4,power_law_1.01,0.8187302398681642
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,2,4,power_law_1.2,0.07424767971038818
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,2,4,power_law_1.2,0.07657343864440917
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,2,4,power_law_1.2,0.0784499216079712
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,2,4,power_law_1.2,0.07710207939147949
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,2,4,power_law_1.2,0.07403520107269287
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,2,4,power_law_1.2,0.07699327945709229
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,2,4,power_law_1.2,0.0798361587524414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,2,4,power_law_1.2,0.08352767944335937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,2,4,power_law_1.2,0.0867750358581543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,2,4,power_law_1.2,0.09169792175292969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,2,4,power_law_1.2,0.10711423873901367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,2,4,power_law_1.2,0.12809215545654298
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,2,4,power_law_1.2,0.1727795219421387
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,2,4,power_law_1.2,0.21889535903930665
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,2,4,power_law_1.2,0.27938816070556644
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,2,4,power_law_1.2,0.04002560138702392
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,2,4,power_law_1.2,0.022082560062408448
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,2,4,power_law_1.2,0.3737587356567383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,2,4,power_law_1.2,0.025345280170440677
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,2,4,power_law_1.2,0.027267839908599857
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,2,4,power_law_1.2,0.03160831928253174
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,2,4,power_law_1.2,0.038407680988311765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,2,4,power_law_1.2,0.045283842086791995
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,2,4,power_law_1.2,0.05188992023468018
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,2,4,power_law_1.2,0.054977278709411624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,2,4,power_law_1.2,0.056883201599121094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,2,4,power_law_1.2,0.05821055889129638
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,2,4,power_law_1.2,0.05494016170501709
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,2,4,power_law_1.2,0.05173759937286378
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,2,4,power_law_1.2,0.05220863819122314
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,2,4,power_law_1.2,0.05516672134399414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,2,4,power_law_1.2,0.5251136016845703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,2,4,power_law_1.2,0.059111680984497074
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,2,4,power_law_1.2,0.06281087875366212
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,2,4,power_law_1.2,0.06700672149658203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,2,4,power_law_1.2,0.07825535774230957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,2,4,power_law_1.2,0.0923583984375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,2,4,power_law_1.2,0.12099583625793457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,2,4,power_law_1.2,0.14398847579956056
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,2,4,power_law_1.2,0.19520383834838867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,2,4,power_law_1.2,0.6925325012207031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,2,4,power_law_1.2,0.24343679428100584
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,1,8,power_law_1.2,1.6994866943359377
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,4,2,balanced,0.017820160388946533
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,4,2,balanced,0.018967039585113525
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,4,2,balanced,0.027731199264526368
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,4,2,balanced,0.03847295999526977
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,4,2,balanced,0.05463935852050781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,4,2,balanced,0.054481921195983884
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,4,2,balanced,0.055406079292297364
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,4,2,balanced,0.05592063903808594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,4,2,balanced,0.06926591873168945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,4,2,balanced,0.07018752098083496
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,2,4,power_law_1.2,0.4395366287231445
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,4,2,balanced,0.07090432167053222
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,4,2,balanced,0.07269631862640381
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,4,2,balanced,0.07350656032562256
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,4,2,balanced,0.07492735862731933
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,4,2,balanced,0.07772543907165527
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,4,2,balanced,0.07899263858795166
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,2,4,power_law_1.2,0.3482150268554688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,4,2,balanced,0.09307007789611817
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,4,2,balanced,0.10318976402282716
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,4,2,balanced,0.13759743690490722
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,4,2,balanced,0.1623859214782715
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,2,4,power_law_1.2,0.9903103637695313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,4,2,balanced,0.23080064773559572
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,4,2,balanced,0.28814975738525395
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,4,2,balanced,0.41737087249755855
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,4,2,balanced,0.0841420841217041
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,4,2,balanced,0.535601921081543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,4,2,balanced,0.7883046722412109
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,4,2,balanced,1.044791030883789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,4,2,balanced,0.020824320316314697
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,4,2,balanced,0.021258239746093748
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,4,2,balanced,0.022874879837036132
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,4,2,balanced,0.031106560230255126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,2,4,power_law_1.2,0.660417251586914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,4,2,balanced,0.0488588809967041
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,4,2,balanced,0.0478937578201294
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,4,2,balanced,0.048368639945983886
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,4,2,balanced,0.049259519577026366
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,4,2,balanced,0.05868800163269043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,1,8,power_law_1.2,0.9958860778808594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,4,2,balanced,0.06509696006774902
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,4,2,balanced,0.06795519828796387
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,4,2,balanced,0.07253759860992431
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,2,4,power_law_1.2,1.2913778686523438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,4,2,balanced,0.07941887855529786
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,4,2,balanced,0.10703871726989747
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,4,2,balanced,0.12333312034606933
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,4,2,balanced,0.18360319137573242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,4,2,balanced,0.2172108840942383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,4,2,balanced,0.3130342483520508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,4,2,balanced,0.4107686233520508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,4,2,balanced,0.06090752124786377
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,4,2,balanced,0.06246912002563476
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,4,2,balanced,0.6088921737670898
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,4,2,power_law_1.01,0.018320640325546263
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,4,2,power_law_1.01,0.026590719223022464
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,4,2,power_law_1.01,0.031134719848632815
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,4,2,balanced,0.06312448024749756
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,4,2,balanced,0.8008076477050782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,4,2,power_law_1.01,0.0365068793296814
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,4,2,power_law_1.01,0.03984256029129028
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,4,2,power_law_1.01,0.05177087783813476
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,4,2,balanced,0.05901311874389649
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,4,2,power_law_1.01,0.06773632049560548
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,4,2,power_law_1.01,0.068122878074646
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,4,2,power_law_1.01,0.06940159797668458
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,4,2,power_law_1.01,0.07126783847808837
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,4,2,power_law_1.01,0.07338624000549317
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,4,2,power_law_1.01,0.07607295989990234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,4,2,power_law_1.01,0.07889791965484619
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,4,2,power_law_1.01,0.08085247993469238
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,4,2,power_law_1.01,0.0871116828918457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,4,2,power_law_1.01,0.09707776069641114
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,4,2,power_law_1.01,0.10934144020080568
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,4,2,power_law_1.01,0.14862848281860352
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,4,2,power_law_1.01,0.05288832187652588
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,4,2,power_law_1.01,0.19553024291992188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,4,2,power_law_1.01,0.2566643142700195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,4,2,power_law_1.01,0.319815673828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,4,2,power_law_1.01,0.44956287384033206
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,4,2,power_law_1.01,0.6054950332641601
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,4,2,power_law_1.01,0.022609920501708985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,4,2,power_law_1.01,0.8687667083740234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,4,2,power_law_1.01,0.053114881515502935
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,4,2,power_law_1.01,0.028847360610961915
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,4,2,power_law_1.01,1.1288345336914063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,4,2,power_law_1.01,0.036607999801635746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,4,2,balanced,0.0876915168762207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,4,2,power_law_1.01,0.04571648120880127
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,4,2,power_law_1.01,0.04704639911651611
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,4,2,balanced,0.06698495864868163
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,4,2,power_law_1.01,0.05482240200042725
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,4,2,power_law_1.01,0.0559603214263916
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,4,2,power_law_1.01,0.057519359588623045
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,4,2,power_law_1.01,0.05889279842376709
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,4,2,power_law_1.01,0.06098176002502441
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,4,2,power_law_1.01,0.08341888427734376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,4,2,power_law_1.01,0.09207167625427246
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,4,2,power_law_1.01,0.030826239585876464
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,4,2,power_law_1.01,0.10154751777648925
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,4,2,power_law_1.01,0.12579456329345703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,4,2,power_law_1.01,0.14933247566223146
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,4,2,power_law_1.01,0.2042176055908203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,4,2,power_law_1.01,0.02076672077178955
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,4,2,power_law_1.01,0.25177215576171874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,4,2,power_law_1.01,0.35912193298339845
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,4,2,power_law_1.01,0.4598976135253906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,4,2,power_law_1.01,0.6677606201171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,4,2,power_law_1.2,0.026877439022064208
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,4,2,power_law_1.2,0.030242559909820554
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,4,2,power_law_1.01,0.8567987060546874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,4,2,power_law_1.2,0.03507200002670288
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,4,2,power_law_1.2,0.04063360214233398
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,4,2,power_law_1.2,0.04975872039794922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,4,2,power_law_1.2,0.05170176029205322
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,4,2,power_law_1.2,0.01817855954170227
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,4,2,power_law_1.2,0.05372288227081299
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,4,2,power_law_1.2,0.06832384109497071
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,4,2,power_law_1.2,0.06824063777923584
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,4,2,power_law_1.01,0.06437119960784912
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,4,2,power_law_1.2,0.06917759895324707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,4,2,power_law_1.01,0.06622591972351075
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,4,2,power_law_1.2,0.07259007930755615
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,4,2,power_law_1.2,0.07328383922576905
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,4,2,power_law_1.01,0.06760575771331787
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,4,2,power_law_1.2,0.07993216037750243
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,4,2,power_law_1.2,0.08774784088134766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,4,2,power_law_1.2,0.09783935546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,4,2,power_law_1.2,0.11110015869140626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,4,2,power_law_1.2,0.1513856029510498
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,4,2,power_law_1.2,0.20029184341430667
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,4,2,power_law_1.2,0.2631308746337891
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,4,2,power_law_1.01,0.047562241554260254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,4,2,power_law_1.2,0.32394622802734374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,4,2,power_law_1.2,0.07583744049072265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,4,2,power_law_1.2,0.08118271827697754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,4,2,power_law_1.2,0.026900479793548582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,4,2,power_law_1.2,0.030344960689544676
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,4,2,power_law_1.2,0.03458303928375244
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,4,2,power_law_1.2,0.04593023777008057
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,4,2,power_law_1.2,0.047202558517456056
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,4,2,power_law_1.2,0.04780543804168701
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,4,2,power_law_1.2,0.05531775951385498
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,4,2,power_law_1.2,0.055900158882141116
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,4,2,power_law_1.2,0.45535617828369135
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,4,2,power_law_1.2,0.057245440483093255
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,4,2,power_law_1.2,0.0621452808380127
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,4,2,power_law_1.2,0.06128255844116211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,4,2,power_law_1.2,0.0677184009552002
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,4,2,power_law_1.2,0.07294976234436035
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,4,2,power_law_1.2,0.08399488449096679
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,4,2,power_law_1.2,0.0940505599975586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,4,2,power_law_1.2,0.10331904411315918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,4,2,power_law_1.2,0.02323456048965454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,4,2,power_law_1.2,0.021049599647521972
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,4,2,power_law_1.2,0.12774271965026857
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,4,2,power_law_1.2,0.15183744430541993
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,4,2,power_law_1.2,0.6338598251342773
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,4,2,power_law_1.2,0.20501375198364258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,2,4,power_law_1.2,0.8656320190429687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,4,2,power_law_1.2,0.25336576461791993
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,1,8,balanced,0.026969599723815917
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,1,8,balanced,0.05037695884704589
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,1,8,balanced,0.06207104206085205
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,1,8,balanced,0.09136511802673339
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,1,8,balanced,0.17340160369873048
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,1,8,balanced,0.22035327911376953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,1,8,balanced,0.22760831832885745
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,1,8,balanced,0.23359231948852538
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,1,8,balanced,0.24257024765014648
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,4,2,power_law_1.2,0.06499711990356445
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,1,8,balanced,0.245034236907959
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,1,8,balanced,0.24047359466552734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,1,8,balanced,0.23974655151367186
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,1,8,balanced,0.24248191833496097
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,1,8,balanced,0.24992000579833987
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,1,8,balanced,0.2612416076660156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,1,8,balanced,0.267509765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,4,2,power_law_1.2,0.896490249633789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,1,8,balanced,0.2866547203063965
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,1,8,balanced,0.3211494445800781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,1,8,balanced,0.35261566162109376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,4,2,power_law_1.2,0.4821696090698242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,1,8,balanced,0.4909312057495117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,4,2,power_law_1.2,0.3695910263061523
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,1,8,balanced,0.6865087890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,1,8,balanced,0.8525862121582032
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,1,8,balanced,0.43491329193115236
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,1,8,balanced,1.2475968170166016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,1,8,balanced,0.02781696081161499
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,1,8,balanced,1.593236541748047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,1,8,balanced,0.03153151988983154
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,1,8,balanced,0.041922559738159185
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,1,8,balanced,0.05921535968780518
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,1,8,balanced,0.10124287605285645
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,1,8,balanced,0.11178751945495606
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,1,8,balanced,0.1157107162475586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,1,8,balanced,0.11934080123901367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,1,8,balanced,0.12356479644775391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,1,8,balanced,2.4027711486816408
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,1,8,balanced,0.12475135803222656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,1,8,balanced,0.11832575798034668
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,1,8,balanced,0.12071423530578614
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,1,8,balanced,0.12337280273437501
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,1,8,balanced,0.12923263549804687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,1,8,balanced,0.13253376007080078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,1,8,balanced,0.13780351638793947
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,1,8,balanced,0.15269375801086427
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,1,8,balanced,0.17498111724853516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,1,8,balanced,0.19249023437499999
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,1,8,balanced,0.22973312377929686
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,1,8,balanced,0.27023103713989255
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,1,8,balanced,0.37189632415771484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,1,8,balanced,0.4506828689575196
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,1,8,balanced,0.6478272247314453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,1,8,power_law_1.01,0.08727295875549317
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,1,8,balanced,3.1831732177734375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,1,8,balanced,1.2231820678710936
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,1,8,power_law_1.01,0.12184063911437988
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,1,8,power_law_1.01,0.1321548843383789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,1,8,power_law_1.01,0.13243776321411133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,1,8,balanced,1.6160716247558593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,1,8,balanced,0.8305420684814454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,1,8,power_law_1.01,0.2075609588623047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,1,8,power_law_1.01,0.20866687774658202
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,1,8,power_law_1.01,0.19522815704345703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,1,8,power_law_1.01,0.10114175796508788
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,1,8,power_law_1.01,0.22356351852416992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,1,8,power_law_1.01,0.23976831436157225
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,1,8,power_law_1.01,0.24063104629516602
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,1,8,power_law_1.01,0.2495731163024902
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,1,8,power_law_1.01,0.26222719192504884
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,1,8,power_law_1.01,0.2649702453613281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,1,8,power_law_1.01,0.2972966384887695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,1,8,power_law_1.01,0.3193728065490723
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,1,8,power_law_1.01,0.38415615081787113
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,1,8,power_law_1.01,0.40728321075439455
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,1,8,power_law_1.01,0.4878105545043946
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,1,8,power_law_1.01,0.6697945404052734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,1,8,power_law_1.01,0.9461529541015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,1,8,power_law_1.01,1.1922431945800782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,1,8,power_law_1.01,0.2134796714782715
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,1,8,power_law_1.01,0.05643775939941407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,1,8,power_law_1.01,1.5710362243652345
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,1,8,power_law_1.01,0.0818393611907959
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,1,8,power_law_1.01,0.07890687942504883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,1,8,power_law_1.01,0.08390399932861328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,1,8,power_law_1.01,0.09464320182800293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,4,2,power_law_1.2,0.6623910522460937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,1,8,power_law_1.01,0.10921600341796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,1,8,power_law_1.01,2.345278778076172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,1,8,power_law_1.01,0.11367424011230469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,1,8,power_law_1.01,0.11623935699462891
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,1,8,power_law_1.01,0.11969663619995116
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,1,8,power_law_1.01,0.12247808456420899
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,1,8,power_law_1.01,0.12430720329284668
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,1,8,power_law_1.01,0.11890432357788086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,1,8,power_law_1.01,0.12280320167541503
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,1,8,power_law_1.01,0.1260927963256836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,1,8,power_law_1.01,0.13348480224609377
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,1,8,power_law_1.01,0.13976832389831545
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,1,8,power_law_1.01,0.10592384338378906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,1,8,power_law_1.01,0.17649152755737305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,1,8,power_law_1.01,0.18913536071777343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,1,8,power_law_1.01,3.4282891845703127
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,1,8,power_law_1.01,0.248590087890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,1,8,power_law_1.01,0.2924287986755371
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,1,8,power_law_1.01,0.40243713378906254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,1,8,power_law_1.01,0.4737715148925782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,1,8,power_law_1.01,0.6810124969482422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,1,8,power_law_1.2,0.08625023841857911
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,1,8,power_law_1.2,0.09507967948913573
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,1,8,power_law_1.01,1.0054310607910157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,1,8,power_law_1.2,0.11463168144226074
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,1,8,power_law_1.01,4.277980041503906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,1,8,power_law_1.2,0.1178662395477295
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,1,8,power_law_1.2,0.19373184204101562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,1,8,power_law_1.2,0.14429951667785645
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,1,8,power_law_1.2,0.20879232406616213
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,1,8,power_law_1.2,0.21166976928710937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,1,8,power_law_1.2,0.21357824325561525
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,1,8,power_law_1.2,0.22547584533691406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,1,8,power_law_1.2,0.2361075210571289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,1,8,power_law_1.2,0.24295936584472658
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,1,8,power_law_1.2,0.2523456001281738
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,1,8,power_law_1.01,1.8661273193359373
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,1,8,power_law_1.2,0.27768320083618164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,1,8,power_law_1.2,0.29384319305419926
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,1,8,power_law_1.2,0.3066470336914063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,1,8,power_law_1.01,1.2819891357421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,1,8,power_law_1.2,0.3515660858154297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,1,8,power_law_1.2,0.4418572616577149
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,1,8,power_law_1.2,0.5976678466796874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,1,8,power_law_1.2,0.751431655883789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,1,8,power_law_1.2,0.3783667373657226
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,1,8,power_law_1.2,1.5320704650878905
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,1,8,power_law_1.2,0.054666237831115724
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,1,8,power_law_1.2,1.0473241424560547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,1,8,power_law_1.2,0.06397312164306641
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,1,8,power_law_1.2,0.07325183868408203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,1,8,power_law_1.2,0.08927359580993652
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,1,8,power_law_1.2,0.08666751861572267
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,1,8,power_law_1.2,0.1076915168762207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,1,8,power_law_1.2,0.10989567756652832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,1,8,power_law_1.2,0.11360256195068359
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,1,8,power_law_1.2,0.11723008155822753
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,1,8,power_law_1.2,0.11783424377441407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,1,8,power_law_1.2,1.9681178283691405
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,1,8,power_law_1.2,0.11755647659301757
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,1,8,power_law_1.2,0.12127743721008302
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,1,8,power_law_1.2,0.1232051181793213
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,1,8,power_law_1.2,0.13214336395263673
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,1,8,power_law_1.2,0.13543807983398437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,1,8,power_law_1.2,0.1479334354400635
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,1,8,power_law_1.2,0.12053503990173339
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,1,8,power_law_1.2,0.1823040008544922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,1,8,power_law_1.2,0.20560640335083008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,1,8,power_law_1.2,0.2702079963684082
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,1,8,power_law_1.2,0.32809726715087895
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,1,8,power_law_1.2,0.4581350326538086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,1,8,power_law_1.2,0.8373632049560547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,1,8,power_law_1.2,0.573262062072754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,4,2,power_law_1.2,1.1492313385009765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,1,8,power_law_1.2,2.5844903564453126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,2,4,balanced,0.0375871992111206
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,2,4,balanced,0.048079361915588384
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,2,4,balanced,0.05916160106658935
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,2,4,balanced,0.09650303840637206
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,2,4,balanced,0.16221439361572265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,2,4,balanced,0.16685312271118163
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,2,4,balanced,0.22095872879028322
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,2,4,balanced,0.22453376770019534
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,2,4,balanced,0.22623487472534182
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,2,4,balanced,0.2320038414001465
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,2,4,balanced,0.22524160385131836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,2,4,balanced,0.23192447662353516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,2,4,balanced,0.23314176559448244
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,2,4,balanced,0.24032512664794922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,2,4,balanced,0.24983808517456058
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,2,4,balanced,0.2549375915527344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,2,4,balanced,0.2642444801330567
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,2,4,balanced,0.3088652801513672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,1,8,power_law_1.2,1.097347869873047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,2,4,balanced,0.41780353546142573
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,2,4,balanced,0.4650688171386719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,2,4,balanced,0.658397445678711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,2,4,balanced,0.33920257568359374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,2,4,balanced,0.8186879730224609
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,2,4,balanced,1.1905535888671874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,2,4,balanced,1.5306355285644533
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,2,4,balanced,0.026807041168212892
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,2,4,balanced,0.030563840866088866
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,2,4,balanced,0.041948161125183105
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,2,4,balanced,0.0625267219543457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,2,4,balanced,0.1020902442932129
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,2,4,balanced,0.10514304161071777
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,2,4,balanced,0.12049152374267577
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,2,4,balanced,0.12297216415405274
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,2,4,balanced,2.2777357482910157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,2,4,balanced,0.1253337574005127
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,2,4,balanced,2.9684121704101565
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,2,4,balanced,0.12793855667114257
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,1,8,power_law_1.2,1.6058560180664063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,2,4,balanced,0.12148096084594726
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,2,4,balanced,0.12010623931884765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,2,4,balanced,0.12990976333618165
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,2,4,balanced,0.1339455986022949
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,2,4,balanced,0.14551424026489257
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,2,4,balanced,0.16601600646972656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,2,4,balanced,0.1800396728515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,2,4,balanced,0.21473535537719726
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,2,4,balanced,0.11815168380737304
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,1,8,power_law_1.2,3.6252236938476563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,2,4,balanced,0.25129087448120113
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,2,4,balanced,0.34787841796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,1,8,power_law_1.2,4.654365539550781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,2,4,balanced,0.41403903961181643
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,2,4,balanced,0.12674816131591798
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,2,4,power_law_1.01,0.05906559944152832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,2,4,balanced,0.6033817672729492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,2,4,power_law_1.01,0.07677567958831787
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,2,4,balanced,0.7691033935546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,2,4,power_law_1.01,0.08864255905151366
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,2,4,power_law_1.01,0.10149375915527344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,2,4,power_law_1.01,0.11273728370666505
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,2,4,power_law_1.01,0.1974220848083496
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,2,4,power_law_1.01,0.20291967391967775
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,2,4,balanced,1.126739196777344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,2,4,power_law_1.01,0.20688896179199218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,2,4,power_law_1.01,0.2109619140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,2,4,power_law_1.01,0.21747840881347655
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,2,4,balanced,1.4821875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,2,4,power_law_1.01,0.21941503524780273
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,2,4,power_law_1.01,0.23695104598999023
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,2,4,power_law_1.01,0.23743743896484376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,2,4,power_law_1.01,0.2622515106201172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,2,4,power_law_1.01,0.27364864349365237
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,2,4,power_law_1.01,0.154269437789917
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,2,4,power_law_1.01,0.35291393280029293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,2,4,power_law_1.01,0.3882099151611328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,2,4,power_law_1.01,0.4663513565063477
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,2,4,power_law_1.01,0.5753177642822266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,2,4,power_law_1.01,0.8383309173583985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,2,4,power_law_1.01,0.2912870407104492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,4,2,power_law_1.2,0.8817062377929688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,2,4,power_law_1.01,1.0229964447021485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,2,4,power_law_1.01,0.04150400161743164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,2,4,power_law_1.01,0.0494374418258667
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,2,4,power_law_1.01,0.05901440143585205
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,2,4,power_law_1.01,1.4425433349609373
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,2,4,power_law_1.01,0.06704127788543701
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,2,4,power_law_1.01,0.08137472152709961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,2,4,power_law_1.01,0.1001471996307373
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,2,4,power_law_1.01,0.11785087585449219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,2,4,power_law_1.01,0.12228992462158203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,2,4,power_law_1.01,0.12341759681701661
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,2,4,power_law_1.01,0.12486144065856934
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,2,4,power_law_1.01,1.9593959045410156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,2,4,power_law_1.01,0.11984640121459962
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,2,4,power_law_1.01,0.12412416458129882
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,2,4,power_law_1.01,0.12852736473083498
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,2,4,power_law_1.01,0.13294464111328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,2,4,power_law_1.01,0.14286208152770996
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,2,4,power_law_1.01,0.16668415069580078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,2,4,power_law_1.01,0.18056320190429687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,2,4,power_law_1.01,0.11925888061523438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,2,4,power_law_1.01,0.24278656005859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,2,4,power_law_1.01,2.7075814819335937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,2,4,power_law_1.01,0.28222719192504886
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,2,4,power_law_1.01,0.1175705623626709
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,2,4,power_law_1.01,0.3676403045654297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,2,4,power_law_1.01,0.49531776428222657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,2,4,power_law_1.2,0.058620162010192864
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,2,4,power_law_1.01,0.686904296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,2,4,power_law_1.2,0.06523647785186767
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,2,4,power_law_1.2,0.08337535858154296
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,2,4,power_law_1.2,0.0973043155670166
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,2,4,power_law_1.01,0.8937433624267579
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,2,4,power_law_1.2,0.10698111534118653
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,2,4,power_law_1.2,0.15227007865905762
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,2,4,power_law_1.2,0.19976959228515626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,2,4,power_law_1.2,0.20862207412719727
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,2,4,power_law_1.2,0.2113804817199707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,2,4,power_law_1.01,1.2489894104003905
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,2,4,power_law_1.2,0.21866111755371093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,2,4,power_law_1.2,0.222869758605957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,2,4,power_law_1.2,0.217891845703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,2,4,power_law_1.2,0.23253631591796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,2,4,power_law_1.2,0.24936960220336912
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,2,4,power_law_1.01,1.6380531311035156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,2,4,power_law_1.2,0.2673113632202148
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,2,4,power_law_1.2,0.27757312774658205
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,2,4,power_law_1.2,0.3137996864318848
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,2,4,power_law_1.2,0.3619276809692383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,2,4,power_law_1.2,0.3940185546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,2,4,power_law_1.01,3.7810983276367187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,1,8,power_law_1.2,2.2676467895507812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,2,4,power_law_1.2,0.4866892623901367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,2,4,power_law_1.2,0.6007487869262695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,2,4,power_law_1.2,0.03995136022567749
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,2,4,power_law_1.2,0.045379838943481444
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,2,4,power_law_1.2,0.8878950500488282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,2,4,power_law_1.2,0.05585792064666748
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,2,4,power_law_1.2,0.0648576021194458
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,2,4,power_law_1.2,0.07155839920043945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,2,4,power_law_1.2,0.10076416015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,2,4,power_law_1.2,0.11653632164001464
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,2,4,power_law_1.2,0.12132863998413086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,2,4,power_law_1.2,0.12313983917236329
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,2,4,power_law_1.2,0.12493311882019043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,2,4,power_law_1.2,1.07936767578125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,2,4,power_law_1.2,0.12632960319519043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,2,4,power_law_1.2,0.11875455856323243
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,2,4,power_law_1.2,0.1205465602874756
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,2,4,power_law_1.2,0.12182271957397461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,2,4,power_law_1.2,0.13438079833984376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,2,4,power_law_1.2,0.14530431747436523
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,2,4,power_law_1.2,0.16713600158691405
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,2,4,power_law_1.2,0.20023168563842772
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,2,4,power_law_1.2,0.2410380744934082
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,2,4,power_law_1.2,0.29755008697509766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,2,4,power_law_1.2,1.50848388671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,2,4,power_law_1.2,0.4029196929931641
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,2,4,power_law_1.2,0.13076607704162596
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,2,4,power_law_1.2,0.7853580474853515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,2,4,power_law_1.2,0.5055219268798827
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,4,2,balanced,0.03764480113983154
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,4,2,balanced,0.04611455917358399
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,4,2,balanced,0.06054912090301514
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,4,2,balanced,0.09293824195861816
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,4,2,balanced,0.15828224182128908
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,4,2,balanced,0.16004608154296876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,4,2,balanced,0.1634444808959961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,4,2,balanced,0.16488832473754883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,4,2,balanced,0.213492488861084
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,4,2,balanced,0.21469823837280275
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,4,2,balanced,0.21495296478271486
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,4,2,balanced,0.21849855422973632
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,4,2,balanced,0.22036991119384766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,2,4,power_law_1.2,1.9987481689453126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,4,2,balanced,0.22415231704711913
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,4,2,balanced,0.22945791244506836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,4,2,balanced,0.23854719161987306
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,4,2,balanced,0.25906431198120117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,4,2,balanced,0.28414207458496094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,4,2,balanced,0.3179507255554199
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,4,2,balanced,0.4023180770874023
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,4,2,balanced,0.4484223937988281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,4,2,balanced,0.6413107299804688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,4,2,balanced,0.8004966735839844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,2,4,power_law_1.2,0.9107456207275391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,4,2,balanced,1.1643993377685546
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,4,2,balanced,0.025470719337463376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,4,2,balanced,0.030145280361175537
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,4,2,balanced,0.043842558860778806
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,4,2,balanced,0.06359168052673339
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,4,2,balanced,0.10732416152954101
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,4,2,balanced,0.108538875579834
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,4,2,balanced,2.2201548767089845
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,4,2,balanced,0.10964223861694336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,4,2,balanced,1.496968994140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,4,2,balanced,0.11147520065307617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,4,2,balanced,0.12840703964233396
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,4,2,balanced,0.1290060806274414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,4,2,balanced,0.1307110404968262
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,4,2,balanced,0.1322111988067627
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,4,2,balanced,0.13357695579528808
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,4,2,balanced,0.1367142391204834
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,4,2,balanced,0.13974911689758301
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,4,2,balanced,2.897864990234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,4,2,balanced,0.14236672401428224
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,4,2,balanced,0.16504447937011718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,4,2,balanced,0.18238208770751954
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,4,2,balanced,0.21113983154296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,4,2,balanced,0.24167552947998047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,4,2,balanced,0.3515647888183594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,4,2,balanced,0.4089715194702149
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,4,2,balanced,0.6000703811645508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,4,2,balanced,0.1504435157775879
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,4,2,power_law_1.01,0.04595327854156494
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,4,2,balanced,0.7687500762939453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,4,2,power_law_1.01,0.054351358413696295
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,4,2,power_law_1.01,0.06717440128326416
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,4,2,balanced,1.1295193481445314
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,4,2,power_law_1.01,0.07957632064819335
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,4,2,power_law_1.01,0.09632767677307129
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,4,2,power_law_1.01,0.1349081611633301
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,4,2,power_law_1.01,0.15440640449523926
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,4,2,balanced,1.4781581115722655
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,4,2,power_law_1.01,0.19537151336669922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,2,4,power_law_1.2,1.3123712158203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,4,2,power_law_1.01,0.19717248916625976
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,4,2,power_law_1.01,0.21128192901611328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,4,2,power_law_1.01,0.21165184020996092
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,4,2,power_law_1.01,0.22990720748901367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,4,2,power_law_1.01,0.24745344161987304
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,4,2,power_law_1.01,0.1506662368774414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,4,2,power_law_1.01,0.2633062362670898
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,4,2,power_law_1.01,0.3436211013793945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,4,2,power_law_1.01,0.3514828872680664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,4,2,power_law_1.01,0.19949440002441404
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,4,2,power_law_1.01,0.42295806884765624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,4,2,power_law_1.01,0.21296768188476561
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,4,2,power_law_1.01,0.5242419052124023
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,4,2,power_law_1.01,0.7327065277099609
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,4,2,power_law_1.01,0.8673677062988281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,4,2,power_law_1.01,0.02679296016693115
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,4,2,power_law_1.01,1.2772198486328126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,4,2,power_law_1.01,0.04096896171569824
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,4,2,power_law_1.01,0.04885119915008545
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,4,2,power_law_1.01,1.6448063659667969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,4,2,power_law_1.01,0.05819519996643067
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,4,2,power_law_1.01,0.06613503932952881
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,4,2,power_law_1.01,0.10045951843261718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,4,2,power_law_1.01,0.10253312110900878
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,4,2,power_law_1.01,0.10506624221801757
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,4,2,power_law_1.01,0.12280447959899903
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,4,2,power_law_1.01,0.12413951873779297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,4,2,power_law_1.01,0.12759296417236327
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,4,2,power_law_1.01,2.3388812255859373
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,4,2,power_law_1.01,0.12949888229370116
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,4,2,power_law_1.01,0.13260031700134278
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,4,2,power_law_1.01,0.13566335678100586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,4,2,power_law_1.01,0.13816703796386717
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,4,2,power_law_1.01,0.1420479965209961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,4,2,power_law_1.01,0.15153663635253906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,4,2,power_law_1.01,0.1721843147277832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,4,2,power_law_1.01,0.191014404296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,4,2,power_law_1.01,0.24816511154174803
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,4,2,power_law_1.01,0.2846860885620117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,4,2,power_law_1.01,3.2743820190429687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,2,4,power_law_1.2,2.845401611328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,4,2,power_law_1.01,0.38403968811035155
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,4,2,power_law_1.01,0.46698623657226557
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,4,2,power_law_1.2,0.0456166410446167
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,4,2,power_law_1.01,0.7042559814453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,4,2,power_law_1.2,0.051363840103149414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,4,2,power_law_1.01,0.8497625732421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,4,2,power_law_1.2,0.06498559951782226
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,4,2,power_law_1.2,0.07675392150878906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,4,2,power_law_1.2,0.09269120216369628
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,4,2,power_law_1.01,1.2039334106445314
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,4,2,power_law_1.2,0.13196800231933595
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,4,2,power_law_1.2,0.15233792304992674
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,4,2,power_law_1.2,0.15799296379089356
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,4,2,power_law_1.2,0.19515647888183593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,4,2,power_law_1.2,0.20438655853271484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,4,2,power_law_1.01,1.578656005859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,4,2,power_law_1.2,0.20724096298217773
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,4,2,power_law_1.2,0.21345535278320313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,4,2,power_law_1.2,0.23220991134643554
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,4,2,power_law_1.2,0.24590335845947267
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,4,2,power_law_1.2,0.2537471961975098
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,4,2,power_law_1.2,0.2736140823364258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,4,2,power_law_1.2,0.34422271728515624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,4,2,power_law_1.2,0.35727745056152344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,4,2,power_law_1.2,0.4314726257324219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,4,2,power_law_1.2,0.19792896270751953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,4,2,power_law_1.2,0.5294054412841798
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,4,2,power_law_1.2,0.7648691558837891
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,4,2,power_law_1.2,0.028718080520629886
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,4,2,power_law_1.2,0.03821439981460571
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,4,2,power_law_1.2,0.8987059020996094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,4,2,power_law_1.2,0.04680448055267334
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,4,2,power_law_1.2,0.05570432186126709
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,4,2,power_law_1.2,0.06607999801635742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,4,2,power_law_1.2,0.10228992462158204
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,4,2,power_law_1.2,0.09790847778320313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,2,4,power_law_1.2,1.8351808166503907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,4,2,power_law_1.2,0.10401408195495605
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,4,2,power_law_1.2,0.12428928375244142
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,4,2,power_law_1.2,0.12604672431945801
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,4,2,power_law_1.2,1.2812492370605468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,4,2,power_law_1.2,0.12698111534118653
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,4,2,power_law_1.2,0.1309235191345215
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,4,2,power_law_1.2,0.13384320259094237
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,4,2,power_law_1.2,0.1375449562072754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,4,2,power_law_1.2,0.13967488288879396
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,4,2,power_law_1.2,0.1432102394104004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,4,2,power_law_1.2,0.15350399971008302
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,2,4,power_law_1.2,4.189204406738281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,4,2,power_law_1.2,0.1921356773376465
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,4,2,power_law_1.2,0.24601343154907224
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,4,2,power_law_1.2,0.28854015350341794
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,4,2,power_law_1.2,0.3978675079345703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,4,2,power_law_1.2,0.17220863342285156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,4,2,power_law_1.2,0.7196428680419922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,8,1,balanced,0.03371903896331787
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,8,1,balanced,0.05012671947479248
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,8,1,balanced,0.0625439977645874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,4,2,power_law_1.2,0.47312255859374996
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,8,1,balanced,0.16443008422851563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,4,2,power_law_1.2,1.6843878173828124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,4,2,power_law_1.2,0.8434611511230468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,8,1,balanced,0.1709996795654297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,8,1,balanced,0.17368064880371095
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,8,1,balanced,0.0957369613647461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,8,1,balanced,0.1824025535583496
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,8,1,balanced,0.18473407745361328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,8,1,balanced,0.18741632461547852
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,8,1,balanced,0.19335872650146485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,8,1,balanced,0.17028032302856447
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,8,1,balanced,0.19703359603881837
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,8,1,balanced,0.19987648010253906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,8,1,balanced,0.20976703643798827
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,8,1,balanced,0.22271488189697264
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,8,1,balanced,0.26994752883911133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,8,1,balanced,0.2982406425476074
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,8,1,balanced,0.47870464324951173
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,8,1,balanced,0.18123199462890624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,8,1,balanced,0.5352166366577149
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,8,1,balanced,0.7884921264648438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,8,1,balanced,0.17629056930541992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,8,1,balanced,1.0029702758789063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,8,1,balanced,1.48112060546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,4,2,power_law_1.2,1.2504998779296874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,8,1,balanced,0.05208960056304932
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,8,1,balanced,1.962895965576172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,8,1,balanced,0.07349376201629639
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,8,1,balanced,0.11415488243103027
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,8,1,balanced,0.11239104270935059
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,8,1,balanced,0.1129689598083496
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,8,1,balanced,0.11472767829895018
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,8,1,balanced,0.11531904220581055
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,8,1,balanced,0.3594611358642578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,8,1,balanced,0.11948351860046387
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,4,2,power_law_1.2,3.3551385498046877
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,8,1,balanced,0.12133119583129884
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,8,1,balanced,0.12415743827819825
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,8,1,balanced,0.12913536071777343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,8,1,balanced,0.1675436782836914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,8,1,balanced,0.16759872436523438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,4,2,power_law_1.2,2.389249267578125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,8,1,balanced,0.1604153633117676
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,8,1,balanced,0.23073087692260744
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,8,1,balanced,0.02969280004501343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,8,1,balanced,0.2803513526916504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,8,1,balanced,0.38441726684570315
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,8,1,balanced,0.03928767919540405
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,8,1,balanced,0.4312294387817383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,8,1,balanced,0.6279379272460938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,8,1,balanced,0.10945023536682127
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,8,1,power_law_1.01,0.03185343980789185
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,8,1,balanced,0.810022430419922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,8,1,balanced,1.2067833709716798
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,8,1,balanced,1.5923712158203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,8,1,power_law_1.01,0.050539522171020514
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,8,1,power_law_1.01,0.1661440086364746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,8,1,power_law_1.01,0.1570841598510742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,8,1,balanced,0.1470041561126709
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,8,1,power_law_1.01,0.16493312835693358
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,8,1,power_law_1.01,0.06204415798187256
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,8,1,power_law_1.01,0.0821286392211914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,4,2,power_law_1.2,1.6361701965332032
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,8,1,balanced,0.17379968643188476
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,8,1,power_law_1.01,0.10050432205200195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,8,1,power_law_1.01,0.22564735412597656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,8,1,power_law_1.01,0.23018495559692384
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,8,1,power_law_1.01,0.1445952033996582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,8,1,power_law_1.01,0.3237363052368164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,8,1,power_law_1.01,0.16608383178710937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,8,1,power_law_1.01,0.3630624008178711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,8,1,power_law_1.01,0.19730047225952146
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,8,1,power_law_1.01,0.3949919891357422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,8,1,power_law_1.01,0.6344550323486329
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,8,1,power_law_1.01,0.2030854415893555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,8,1,power_law_1.01,0.21049087524414062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,8,1,power_law_1.01,0.2195769691467285
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,8,1,power_law_1.01,1.1050822448730468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,8,1,power_law_1.01,0.04097792148590088
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,8,1,power_law_1.01,0.051904640197753905
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,8,1,power_law_1.01,0.06269184112548828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,8,1,power_law_1.01,0.2541235160827637
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,8,1,power_law_1.01,0.07249152183532714
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,8,1,power_law_1.01,2.0532908630371094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,8,1,power_law_1.01,0.09834112167358398
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,8,1,power_law_1.01,0.10426176071166993
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,8,1,power_law_1.01,0.10729984283447265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,8,1,power_law_1.01,0.11169024467468262
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,8,1,power_law_1.01,1.5811013793945312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,8,1,power_law_1.01,0.12051008224487306
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,8,1,power_law_1.01,0.24044672012329102
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,8,1,power_law_1.01,0.1257222366333008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,8,1,power_law_1.01,0.13575360298156738
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,8,1,power_law_1.01,0.15090432167053222
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,8,1,power_law_1.01,0.1607040023803711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,8,1,power_law_1.01,0.1710483169555664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,8,1,power_law_1.01,0.5249708938598633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,8,1,power_law_1.01,0.18876415252685547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,8,1,power_law_1.01,0.20475839614868163
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,8,1,power_law_1.01,0.29764928817749026
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,8,1,power_law_1.01,0.4165708923339844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,8,1,power_law_1.01,0.4981862258911133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,8,1,power_law_1.01,0.8815193939208985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,8,1,power_law_1.01,0.6934944152832031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,8,1,power_law_1.01,0.02954432010650635
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,8,1,power_law_1.2,0.033925759792327884
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,8,1,power_law_1.01,0.8811788940429688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,8,1,power_law_1.2,0.050957441329956055
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,8,1,power_law_1.2,0.060453119277954104
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,8,1,power_law_1.01,0.1102188777923584
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,8,1,power_law_1.2,0.07745279788970946
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,8,1,power_law_1.01,1.274938201904297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,8,1,power_law_1.2,0.1465177631378174
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,8,1,power_law_1.2,0.16497024536132812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,8,1,power_law_1.01,1.7137619018554688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,8,1,power_law_1.2,0.2086227226257324
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,8,1,power_law_1.2,0.21382911682128905
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,8,1,power_law_1.01,0.1684102439880371
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,8,1,power_law_1.2,0.22545791625976563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,8,1,power_law_1.2,0.16739776611328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,8,1,power_law_1.2,0.23113727569580078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,8,1,power_law_1.2,0.24075136184692383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,8,1,power_law_1.2,0.25602943420410157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,8,1,power_law_1.2,0.09477760314941405
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,8,1,power_law_1.2,0.3296108627319336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,8,1,power_law_1.01,0.27800960540771485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,8,1,power_law_1.2,0.15987071990966797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,8,1,power_law_1.2,0.38069759368896483
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,8,1,power_law_1.2,0.1683283233642578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,8,1,power_law_1.2,0.19806272506713868
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,8,1,power_law_1.2,0.6490451049804687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,8,1,power_law_1.2,0.22126399993896487
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,8,1,power_law_1.2,0.04150271892547607
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,8,1,power_law_1.2,0.8790303802490234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,8,1,power_law_1.2,0.051520638465881355
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,8,1,power_law_1.2,0.39664703369140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,8,1,power_law_1.2,0.06075007915496826
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,8,1,power_law_1.2,0.06923711776733399
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,8,1,power_law_1.2,0.10193216323852539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,8,1,power_law_1.2,0.5308915328979492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,8,1,power_law_1.2,0.10520832061767578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,8,1,power_law_1.2,0.10869695663452147
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,8,1,power_law_1.2,0.11181887626647949
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,8,1,power_law_1.2,0.11376128196716309
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,8,1,power_law_1.2,0.1206118392944336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,8,1,power_law_1.2,0.1278540802001953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,8,1,power_law_1.2,0.13905407905578612
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,8,1,power_law_1.2,0.15210240364074706
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,8,1,power_law_1.2,0.030144639015197754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,8,1,power_law_1.2,0.16249343872070313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,8,1,power_law_1.2,0.17220479965209962
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,8,1,power_law_1.2,0.16938175201416014
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,8,1,power_law_1.2,0.19110847473144532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,8,1,power_law_1.2,0.5092486572265625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,8,1,power_law_1.2,1.1088140869140626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,8,1,power_law_1.2,0.2990559959411621
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,8,1,power_law_1.2,0.28632959365844723
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,8,1,power_law_1.2,0.20725631713867188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,1,8,balanced,0.06805632114410401
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,1,8,balanced,0.07925759792327881
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,8,1,power_law_1.2,0.7013804626464843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,8,1,power_law_1.2,0.41738304138183596
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,1,8,balanced,0.11277055740356445
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,1,8,balanced,0.2116620826721191
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,1,8,balanced,0.3883776092529297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,1,8,balanced,0.7195635223388672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,1,8,balanced,0.7267699432373047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,1,8,balanced,0.7292275238037109
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,1,8,balanced,0.7361497497558593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,1,8,balanced,0.7390975952148438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,1,8,balanced,0.7322803497314453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,1,8,balanced,0.7471539306640624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,1,8,balanced,0.7617842864990234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,1,8,balanced,0.773443832397461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,1,8,balanced,0.7920499420166015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,1,8,balanced,0.811299819946289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,1,8,balanced,0.8311628723144532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,1,8,balanced,0.8829299163818358
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,1,8,balanced,0.9138956451416016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,1,8,balanced,1.089876480102539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,8,1,power_law_1.2,0.8853612518310546
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,1,8,balanced,1.1889881896972656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,1,8,balanced,1.6726860046386718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,1,8,balanced,2.0821466064453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,1,8,balanced,2.8564108276367186
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,1,8,balanced,0.04423168182373047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,1,8,balanced,0.05042304039001465
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,1,8,balanced,0.06924799919128419
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,1,8,balanced,3.8837326049804686
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,1,8,balanced,0.12473983764648437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,1,8,balanced,0.2092492866516113
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,1,8,balanced,0.28905855178833006
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,1,8,balanced,0.29684480667114255
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,1,8,balanced,0.2998399925231934
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,1,8,balanced,0.3022067260742188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,1,8,balanced,0.3058700752258301
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,1,8,balanced,0.29799680709838866
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,1,8,balanced,0.3026815986633301
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,1,8,balanced,0.3051456069946289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,1,8,balanced,0.30993663787841796
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,1,8,balanced,5.769987792968751
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,1,8,balanced,0.31948543548583985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,1,8,balanced,0.3251571273803711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,1,8,balanced,0.3434239959716797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,1,8,balanced,0.3774732971191406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,1,8,balanced,0.39937023162841795
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,1,8,balanced,0.4576089477539062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,1,8,balanced,0.515609588623047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,8,1,power_law_1.2,1.5857568359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,1,8,balanced,0.7270464324951172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,1,8,balanced,0.8580274963378907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,1,8,balanced,1.2042060852050782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,1,8,balanced,7.978120727539062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,1,8,power_law_1.01,0.19176704406738282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,1,8,balanced,1.5747059631347657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,1,8,power_law_1.01,0.3667596817016602
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,1,8,power_law_1.01,0.3133452796936035
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,1,8,power_law_1.01,0.3800640106201172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,1,8,balanced,2.163544311523437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,1,8,power_law_1.01,0.4118438339233398
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,1,8,power_law_1.01,0.6597682952880859
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,1,8,power_law_1.01,0.6872601318359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,1,8,power_law_1.01,0.6830028533935547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,1,8,power_law_1.01,0.6949964904785156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,1,8,balanced,2.9101300048828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,1,8,power_law_1.01,0.7009945678710937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,1,8,power_law_1.01,0.726635513305664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,1,8,power_law_1.01,0.7323609924316405
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,1,8,power_law_1.01,0.7602162933349609
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,1,8,power_law_1.01,0.7667520141601563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,1,8,power_law_1.01,0.8031053161621093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,1,8,power_law_1.01,0.824474868774414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,1,8,power_law_1.01,0.9004953765869141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,1,8,power_law_1.01,0.9731929779052735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,1,8,power_law_1.01,1.0911846160888672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,1,8,power_law_1.01,1.3023385620117187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,1,8,power_law_1.01,1.6459097290039062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,1,8,power_law_1.01,2.249845733642578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,1,8,power_law_1.01,2.7340097045898437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,1,8,power_law_1.01,0.1251353645324707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,1,8,power_law_1.01,0.18339712142944337
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,1,8,power_law_1.01,0.16079744338989257
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,1,8,power_law_1.01,0.18252031326293944
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,1,8,power_law_1.01,4.399009399414062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,1,8,power_law_1.01,0.16165760040283203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,1,8,power_law_1.01,0.25103872299194335
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,1,8,power_law_1.01,0.25151615142822265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,1,8,power_law_1.01,0.26076032638549806
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,1,8,power_law_1.01,0.263317756652832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,1,8,power_law_1.01,0.2667212867736816
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,1,8,power_law_1.01,5.3163824462890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,1,8,power_law_1.01,0.2826995277404785
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,1,8,power_law_1.01,0.279422721862793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,8,1,power_law_1.2,1.2723852539062501
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,1,8,power_law_1.01,0.2844812774658203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,1,8,power_law_1.01,0.28016384124755855
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,1,8,power_law_1.01,0.29825279235839847
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,1,8,power_law_1.01,0.3049920082092285
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,1,8,power_law_1.01,0.355676155090332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,1,8,power_law_1.01,0.4037670516967774
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,1,8,power_law_1.01,0.47817214965820315
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,1,8,power_law_1.01,0.5723123168945312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,1,8,power_law_1.01,0.3248051071166992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,1,8,power_law_1.01,0.7909490966796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,1,8,power_law_1.01,0.9428454589843749
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,1,8,power_law_1.01,8.406214599609374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,1,8,power_law_1.01,1.474694366455078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,1,8,power_law_1.2,0.20313215255737305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,1,8,power_law_1.01,1.7370022583007814
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,1,8,power_law_1.2,0.2508108711242676
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,1,8,power_law_1.2,0.3579520034790039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,1,8,power_law_1.2,0.37817470550537113
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,1,8,power_law_1.2,0.44853374481201175
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,1,8,power_law_1.01,2.5504908752441406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,1,8,power_law_1.2,0.664238052368164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,1,8,power_law_1.2,0.6802534484863282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,1,8,power_law_1.2,0.6647872161865235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,1,8,power_law_1.01,9.808998413085938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,1,8,power_law_1.2,0.7034406280517578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,1,8,power_law_1.2,0.7151705932617187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,1,8,power_law_1.2,0.7392281341552734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,1,8,power_law_1.2,0.7456409454345703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,1,8,power_law_1.2,0.7706278228759766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,1,8,power_law_1.01,3.850863342285156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,1,8,power_law_1.2,0.8088832092285155
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,1,8,power_law_1.2,0.8352025604248047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,1,8,power_law_1.2,0.8642098999023438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,1,8,power_law_1.2,0.9208716583251952
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,1,8,power_law_1.2,1.058887710571289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,1,8,power_law_1.2,1.1322930908203124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,1,8,power_law_1.2,1.4084556579589844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,1,8,power_law_1.2,1.8378060913085938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,1,8,power_law_1.2,2.308107452392578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,1,8,power_law_1.2,0.12535807609558106
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,1,8,power_law_1.2,2.8944525146484374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,1,8,power_law_1.2,0.1312012767791748
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,8,1,power_law_1.2,2.0829656982421874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,1,8,power_law_1.2,0.17907072067260743
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,1,8,power_law_1.2,0.1924095916748047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,1,8,power_law_1.2,0.2268889617919922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,1,8,power_law_1.2,0.2330099105834961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,1,8,power_law_1.2,0.24744447708129882
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,1,8,power_law_1.2,0.2594175910949707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,1,8,power_law_1.2,0.2643187141418457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,1,8,power_law_1.2,0.27663999557495117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,1,8,power_law_1.2,0.2941196823120117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,1,8,power_law_1.2,0.2910592079162598
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,1,8,power_law_1.2,0.29085311889648435
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,1,8,power_law_1.2,0.3023500823974609
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,1,8,power_law_1.2,0.3104179191589355
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,1,8,power_law_1.2,0.3214796829223633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,1,8,power_law_1.2,4.187833557128906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,1,8,power_law_1.2,0.32874881744384765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,1,8,power_law_1.2,0.36776702880859374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,1,8,power_law_1.2,0.419947509765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,1,8,power_law_1.2,0.5121664047241211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,1,8,power_law_1.2,0.6562060546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,1,8,power_law_1.2,6.170870971679688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,1,8,power_law_1.2,0.8199219512939454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,1,8,power_law_1.2,1.17683837890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,2,4,balanced,0.06441088199615479
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,2,4,balanced,0.07914239883422851
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,1,8,power_law_1.2,1.5707916259765624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,2,4,balanced,0.11447039604187012
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,2,4,balanced,0.19854848861694335
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,2,4,balanced,0.38436737060546877
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,1,8,power_law_1.2,2.040552978515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,2,4,balanced,0.47876735687255856
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,2,4,balanced,0.7039449310302734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,2,4,balanced,0.7135282897949219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,2,4,balanced,0.7141849517822265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,2,4,balanced,0.7203763580322265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,2,4,balanced,0.7181209564208985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,2,4,balanced,0.726513900756836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,2,4,balanced,0.7379596710205079
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,1,8,power_law_1.2,8.478316650390624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,2,4,balanced,0.753834228515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,2,4,balanced,0.7676313781738282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,1,8,power_law_1.2,2.8328973388671876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,2,4,balanced,0.7901849365234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,2,4,balanced,0.8046886444091796
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,2,4,balanced,0.8586252593994141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,2,4,balanced,0.8980249786376954
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,2,4,balanced,1.0811622619628907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,2,4,balanced,1.1740518188476563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,2,4,balanced,1.6262451171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,2,4,balanced,2.038269500732422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,2,4,balanced,0.042558717727661136
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,2,4,balanced,0.050704641342163084
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,2,4,balanced,0.07385216236114503
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,2,4,balanced,2.7754483032226562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,8,1,power_law_1.2,1.7026739501953123
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,2,4,balanced,0.1273472023010254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,1,8,power_law_1.2,4.233729248046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,2,4,balanced,0.21533439636230467
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,1,8,power_law_1.2,12.513543701171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,2,4,balanced,0.2637273597717285
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,2,4,balanced,0.30070528030395505
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,2,4,balanced,0.3116172790527344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,2,4,balanced,0.30634368896484376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,2,4,balanced,0.3188582420349121
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,2,4,balanced,0.2988863945007324
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,2,4,balanced,0.2997452735900879
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,2,4,balanced,3.714967041015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,2,4,balanced,0.3063347244262695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,2,4,balanced,0.30760576248168947
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,2,4,balanced,0.3167398452758789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,2,4,balanced,0.3197439956665039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,2,4,balanced,0.329986572265625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,2,4,balanced,0.3722342300415039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,2,4,balanced,0.38345214843750003
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,2,4,balanced,0.43372287750244143
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,2,4,balanced,0.4957555389404297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,2,4,balanced,0.6909311676025391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,2,4,balanced,0.7926182556152344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,2,4,balanced,5.266254272460937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,2,4,power_law_1.01,0.10987135887145996
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,2,4,power_law_1.01,0.17529216766357422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,2,4,balanced,1.1330150604248046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,2,4,power_law_1.01,0.1813657569885254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,2,4,power_law_1.01,0.27570816040039064
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,2,4,power_law_1.01,0.2976140785217285
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,2,4,balanced,1.4822515869140624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,2,4,power_law_1.01,0.3851839828491211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,2,4,power_law_1.01,0.6573401641845703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,2,4,power_law_1.01,0.6551590728759765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,2,4,balanced,2.026466522216797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,2,4,power_law_1.01,0.6823283386230469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,2,4,power_law_1.01,0.6824652862548828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,2,4,power_law_1.01,0.7027967834472657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,2,4,power_law_1.01,0.7069324493408203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,2,4,power_law_1.01,0.7256678771972657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,2,4,balanced,2.7074151611328126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,2,4,power_law_1.01,0.753918685913086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,2,4,power_law_1.01,0.7818943786621093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,2,4,power_law_1.01,0.8092082977294922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,2,4,power_law_1.01,0.8448537445068359
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,2,4,power_law_1.01,0.9338419342041016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,2,4,power_law_1.01,0.9938329315185547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,2,4,power_law_1.01,1.1748108673095703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,2,4,power_law_1.01,1.4090559387207031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,2,4,power_law_1.01,0.07234432220458983
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,2,4,power_law_1.01,1.9641331481933595
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,2,4,power_law_1.01,0.11088512420654298
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,2,4,power_law_1.01,0.11321984291076662
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,2,4,power_law_1.01,2.3345587158203123
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,2,4,power_law_1.01,0.13765760421752932
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,2,4,power_law_1.01,0.15077631950378417
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,2,4,power_law_1.01,0.2267302322387695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,2,4,power_law_1.01,0.24871040344238282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,2,4,power_law_1.01,0.2548467254638672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,2,4,power_law_1.01,0.25936384201049806
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,2,4,power_law_1.01,0.26320255279541016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,2,4,power_law_1.01,3.4015512084960937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,2,4,balanced,6.928725585937499
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,2,4,power_law_1.01,0.2689536094665527
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,2,4,power_law_1.01,0.264770565032959
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,2,4,power_law_1.01,0.26672384262084964
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,2,4,power_law_1.01,0.2922406387329102
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,2,4,power_law_1.01,0.29116287231445315
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,2,4,power_law_1.01,0.3012518310546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,2,4,power_law_1.01,0.3197849655151367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,2,4,power_law_1.01,0.3630553436279297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,2,4,power_law_1.01,0.38745471954345706
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,2,4,power_law_1.01,4.323565979003907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,2,4,power_law_1.01,0.5301004791259766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,2,4,power_law_1.01,0.7058560180664062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,2,4,power_law_1.01,0.8803008270263671
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,2,4,power_law_1.2,0.11098239898681642
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,2,4,power_law_1.01,0.4657587051391602
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,2,4,power_law_1.01,1.3022476196289063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,2,4,power_law_1.2,0.1410854434967041
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,2,4,power_law_1.2,0.18882688522338867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,2,4,power_law_1.01,1.6678451538085937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,2,4,power_law_1.2,0.25410560607910154
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,2,4,power_law_1.01,6.1798974609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,2,4,power_law_1.2,0.3257279968261719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,2,4,power_law_1.2,0.3955136108398437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,2,4,power_law_1.01,2.248668212890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,2,4,power_law_1.2,0.6603187561035156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,2,4,power_law_1.2,0.6638464355468751
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,2,4,power_law_1.2,0.6783232116699219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,2,4,power_law_1.2,0.6976780700683594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,2,4,power_law_1.2,0.7235404968261718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,2,4,power_law_1.2,0.7308659362792969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,2,4,power_law_1.01,3.0052786254882813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,2,4,power_law_1.2,0.746138916015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,2,4,power_law_1.2,0.7626048278808594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,2,4,power_law_1.2,0.7878221130371094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,2,4,power_law_1.01,8.483176879882812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,2,4,power_law_1.2,0.7796083068847656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,2,4,power_law_1.2,0.8454835510253906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,2,4,power_law_1.2,0.9360614776611328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,2,4,power_law_1.2,1.208485107421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,2,4,power_law_1.2,1.4784332275390626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,2,4,power_law_1.2,1.0170995330810546
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,2,4,power_law_1.2,0.07174528121948243
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,2,4,power_law_1.2,0.08756863594055175
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,2,4,power_law_1.2,0.11525504112243652
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,2,4,power_law_1.2,2.058693084716797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,2,4,power_law_1.2,0.13723520278930662
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,2,4,power_law_1.2,2.517580871582031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,2,4,power_law_1.2,0.15126144409179687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,2,4,power_law_1.2,0.2296294403076172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,2,4,power_law_1.2,0.25842048645019533
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,2,4,power_law_1.2,0.26079103469848636
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,2,4,power_law_1.2,0.2744998359680176
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,2,4,power_law_1.2,0.2854924774169922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,2,4,power_law_1.2,3.8480090332031254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,2,4,power_law_1.2,0.28076160430908204
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,2,4,power_law_1.2,0.2949171257019043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,2,4,power_law_1.2,0.2478118324279785
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,2,4,power_law_1.2,0.3042495918273926
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,2,4,power_law_1.2,0.31907072067260744
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,2,4,power_law_1.2,0.3218534469604492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,2,4,power_law_1.2,0.36412158966064456
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,2,4,power_law_1.2,0.2778432083129883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,2,4,power_law_1.2,0.3955801773071289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,2,4,power_law_1.2,0.486827507019043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,2,4,power_law_1.2,4.5143194580078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,2,4,power_law_1.2,0.545420799255371
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,2,4,power_law_1.2,0.7240831756591797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,4,2,balanced,0.059552001953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,2,4,power_law_1.2,0.9576038360595703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,4,2,balanced,0.07405183792114259
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,4,2,balanced,0.11209728240966796
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,4,2,balanced,0.19366783142089844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,2,4,power_law_1.2,1.3253631591796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,4,2,balanced,0.3833766555786133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,2,4,power_law_1.2,1.705382385253906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,4,2,balanced,0.47219200134277345
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,4,2,balanced,0.47758975982666013
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,4,2,balanced,0.476519660949707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,4,2,balanced,0.4809894561767578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,4,2,balanced,0.6933875274658203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,4,2,balanced,0.6991270446777345
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,4,2,balanced,0.7072870635986328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,4,2,balanced,0.7181414031982423
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,4,2,balanced,0.7313817596435547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,4,2,balanced,0.7459008026123046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,4,2,balanced,0.7680409240722657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,4,2,balanced,0.7978765106201171
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,4,2,balanced,0.8335091400146485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,4,2,balanced,0.8662028503417968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,4,2,balanced,1.0782617950439453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,4,2,balanced,1.1655206298828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,4,2,balanced,1.6105087280273438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,4,2,balanced,2.0324557495117186
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,4,2,balanced,2.7247271728515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,4,2,balanced,0.039851520061492916
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,4,2,balanced,0.051581439971923826
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,4,2,balanced,0.07180031776428222
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,2,4,power_law_1.2,3.4790093994140627
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,4,2,balanced,3.606106872558594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,4,2,balanced,0.12825087547302244
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,4,2,balanced,0.22118656158447267
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,4,2,balanced,0.2714752006530762
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,4,2,balanced,0.26948095321655274
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,4,2,balanced,0.27254400253295896
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,4,2,balanced,0.2733299255371094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,4,2,balanced,0.31264127731323244
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,4,2,balanced,0.31591167449951174
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,4,2,balanced,0.31675392150878906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,4,2,balanced,0.3195008087158203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,2,4,power_law_1.2,6.92481689453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,2,4,power_law_1.2,2.542078704833984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,4,2,balanced,0.32336769104003904
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,4,2,balanced,0.3269171142578125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,4,2,balanced,0.33202816009521485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,4,2,balanced,0.34479488372802736
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,4,2,balanced,5.134599609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,4,2,balanced,0.3717388916015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,4,2,balanced,0.39354366302490235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,2,4,power_law_1.2,9.464285278320313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,4,2,balanced,0.4270668792724609
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,4,2,balanced,0.4773376083374023
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,4,2,balanced,0.7030220794677734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,4,2,balanced,0.7978457641601563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,4,2,power_law_1.01,0.07449728012084961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,4,2,balanced,1.1497971343994142
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,4,2,power_law_1.01,0.10490240097045897
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,4,2,power_law_1.01,0.12997504234313967
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,4,2,power_law_1.01,0.1868659210205078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,4,2,power_law_1.01,0.2984576034545898
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,4,2,balanced,1.5969343566894532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,4,2,balanced,6.701450195312501
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,4,2,power_law_1.01,0.4011840057373047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,4,2,power_law_1.01,0.4243366241455078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,4,2,power_law_1.01,0.4195852661132813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,4,2,balanced,2.0207078552246096
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,4,2,power_law_1.01,0.4467379379272461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,4,2,power_law_1.01,0.6590656280517578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,4,2,power_law_1.01,0.6753421020507813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,4,2,power_law_1.01,0.6831692504882813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,4,2,balanced,2.711229553222656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,4,2,power_law_1.01,0.7268940734863281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,4,2,power_law_1.01,0.7438169860839844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,4,2,power_law_1.01,0.7574809265136719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,4,2,power_law_1.01,0.798703384399414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,4,2,power_law_1.01,0.8861209869384765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,4,2,power_law_1.01,0.9298009490966797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,4,2,power_law_1.01,0.6951590728759766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,4,2,power_law_1.01,1.2931980895996094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,4,2,power_law_1.01,0.05143295764923096
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,4,2,power_law_1.01,1.7233984374999998
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,4,2,power_law_1.01,0.0680294418334961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,4,2,power_law_1.01,1.0859302520751952
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,4,2,power_law_1.01,2.110967102050781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,4,2,power_law_1.01,0.1215334415435791
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,4,2,power_law_1.01,0.15300607681274414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,4,2,power_law_1.01,0.2244710350036621
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,4,2,power_law_1.01,0.22434047698974607
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,4,2,power_law_1.01,0.22927360534667968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,4,2,power_law_1.01,2.912116394042969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,4,2,power_law_1.01,0.23807231903076173
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,4,2,power_law_1.01,0.26521600723266603
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,4,2,power_law_1.01,0.08521727561950684
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,4,2,power_law_1.01,0.2733721542358398
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,4,2,power_law_1.01,0.28440704345703127
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,4,2,power_law_1.01,0.2800115203857422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,4,2,power_law_1.01,0.2978534317016602
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,4,2,power_law_1.01,0.3068147277832031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,4,2,power_law_1.01,3.7084658813476565
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,4,2,power_law_1.01,0.33134719848632815
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,4,2,power_law_1.01,0.3670054244995117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,4,2,power_law_1.01,0.39220481872558594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,4,2,power_law_1.01,0.4709606552124024
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,4,2,power_law_1.01,0.5395455932617187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,4,2,power_law_1.01,0.29434368133544925
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,4,2,power_law_1.01,0.6769856262207031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,4,2,power_law_1.01,0.886255340576172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,4,2,power_law_1.2,0.07582335948944093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,4,2,power_law_1.2,0.09618687629699707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,4,2,power_law_1.01,1.2015398406982423
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,4,2,power_law_1.01,5.3810357666015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,4,2,power_law_1.2,0.17341184616088867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,4,2,power_law_1.2,0.23971071243286132
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,4,2,power_law_1.01,1.524285430908203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,4,2,power_law_1.2,0.39900543212890627
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,4,2,power_law_1.2,0.4233792114257812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,4,2,power_law_1.2,0.41921535491943357
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,4,2,power_law_1.2,0.4329792022705078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,4,2,power_law_1.2,0.13628543853759764
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,4,2,power_law_1.2,0.6613875579833984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,4,2,power_law_1.01,2.266314239501953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,4,2,power_law_1.2,0.6759718322753907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,4,2,power_law_1.2,0.6937779235839844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,4,2,power_law_1.2,0.7124249267578124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,4,2,power_law_1.01,7.22271484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,4,2,power_law_1.2,0.7172505950927734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,4,2,power_law_1.01,2.9208538818359377
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,4,2,power_law_1.2,0.7271257781982422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,4,2,power_law_1.2,0.7716070556640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,4,2,power_law_1.2,0.7887155151367187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,4,2,power_law_1.2,0.9003865814208984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,4,2,power_law_1.2,0.9349696350097656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,4,2,power_law_1.2,1.1155430603027345
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,4,2,power_law_1.2,1.3068479919433593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,4,2,power_law_1.2,0.05067520141601563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,4,2,power_law_1.2,0.06300159931182861
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,4,2,power_law_1.2,0.08513792037963867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,4,2,power_law_1.2,0.11610367774963379
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,4,2,power_law_1.2,0.13986047744750976
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,4,2,power_law_1.2,2.230243835449219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,4,2,power_law_1.2,0.2150604820251465
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,4,2,power_law_1.2,0.22544767379760744
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,4,2,power_law_1.2,0.23496063232421874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,4,2,power_law_1.2,0.23726335525512696
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,4,2,power_law_1.2,1.6828889465332033
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,4,2,power_law_1.2,0.27547264099121094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,4,2,power_law_1.2,3.007805480957031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,4,2,power_law_1.2,0.283874568939209
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,4,2,power_law_1.2,0.28509056091308593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,4,2,power_law_1.2,0.29288448333740236
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,4,2,power_law_1.2,0.2988863945007324
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,4,2,power_law_1.2,0.30856191635131835
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,4,2,power_law_1.2,0.31882368087768553
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,4,2,power_law_1.2,0.33671680450439456
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,4,2,power_law_1.2,0.3684659194946289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,4,2,power_law_1.2,0.4013644790649414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,4,2,power_law_1.2,0.4877555084228516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,4,2,power_law_1.2,3.8639297485351562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,4,2,power_law_1.2,0.5644236755371094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,4,2,power_law_1.2,0.7157055664062499
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,4,2,power_law_1.2,0.8870976257324219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,4,2,power_law_1.2,1.229153289794922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,8,1,balanced,0.07011136054992675
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,8,1,balanced,0.08702783584594727
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,8,1,balanced,0.12197504043579102
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,8,1,balanced,0.2025574493408203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,8,1,balanced,0.395109748840332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,4,2,power_law_1.2,1.6180493164062502
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,8,1,balanced,0.48353664398193363
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,8,1,balanced,0.4857542419433594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,8,1,balanced,0.4897926330566406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,4,2,power_law_1.2,2.403950042724609
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,8,1,balanced,0.4933350372314453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,8,1,balanced,0.49585025787353515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,8,1,balanced,0.5013222503662109
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,8,1,balanced,0.5046918487548828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,8,1,balanced,0.7179366302490234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,8,1,balanced,0.7266118621826172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,8,1,balanced,0.7327200317382813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,8,1,balanced,0.7528166198730469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,8,1,balanced,0.7811334228515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,8,1,balanced,0.80451904296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,8,1,balanced,0.8412921905517579
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,8,1,balanced,1.1949120330810548
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,8,1,balanced,1.210302734375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,8,1,balanced,1.5959890747070313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,8,1,balanced,2.057489318847656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,8,1,balanced,0.0524889612197876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,8,1,balanced,0.063787522315979
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,8,1,balanced,2.7112332153320313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,4,2,power_law_1.2,3.0543295288085934
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,8,1,balanced,0.08982336044311523
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,8,1,balanced,0.15901247978210448
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,8,1,balanced,0.31960832595825195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,4,2,power_law_1.2,7.475452270507813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,8,1,balanced,0.3236665725708008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,4,2,power_law_1.2,5.596185302734375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,8,1,balanced,0.3232953643798828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,8,1,balanced,0.32706111907958985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,8,1,balanced,0.32525184631347653
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,8,1,balanced,3.6343591308593752
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,8,1,balanced,0.3308396911621094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,8,1,balanced,0.39417343139648436
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,8,1,balanced,0.39797695159912105
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,8,1,balanced,0.40452224731445313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,8,1,balanced,0.40462783813476566
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,8,1,balanced,0.4116960144042968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,8,1,balanced,0.432017936706543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,8,1,balanced,0.45155902862548825
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,8,1,balanced,0.4957542419433594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,8,1,balanced,0.5445017623901367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,8,1,balanced,5.202601318359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,8,1,balanced,0.86774658203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,8,1,balanced,0.9999641418457031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,8,1,balanced,0.2563046455383301
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,8,1,power_law_1.01,0.07144256114959717
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,8,1,power_law_1.01,0.0860313606262207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,8,1,balanced,1.4055622863769532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,8,1,balanced,0.32888446807861327
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,8,1,power_law_1.01,0.12148223876953126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,8,1,power_law_1.01,0.17481407165527343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,8,1,power_law_1.01,0.24289663314819335
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,8,1,balanced,1.7957714843750001
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,8,1,power_law_1.01,0.39510398864746094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,8,1,power_law_1.01,0.41165439605712895
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,8,1,power_law_1.01,0.42577472686767576
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,8,1,power_law_1.01,0.4400339126586914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,8,1,balanced,6.951036376953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,8,1,power_law_1.01,0.45022399902343746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,8,1,power_law_1.01,0.49357631683349606
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,8,1,power_law_1.01,0.5239571380615234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,8,1,balanced,2.627400207519531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,8,1,power_law_1.01,0.6841779327392578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,8,1,power_law_1.01,0.7085638427734375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,8,1,power_law_1.01,0.7297331237792969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,8,1,power_law_1.01,0.7225132751464843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,8,1,power_law_1.01,0.7667378997802734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,8,1,power_law_1.01,0.8531231689453126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,8,1,balanced,3.3619699096679687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,8,1,power_law_1.01,0.9131021118164062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,8,1,power_law_1.01,1.0301459503173827
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,8,1,power_law_1.01,1.217928924560547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,8,1,power_law_1.01,0.06418176174163819
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,8,1,power_law_1.01,1.5742387390136718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,8,1,power_law_1.01,0.08949503898620606
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,8,1,power_law_1.01,1.9545983886718752
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,8,1,power_law_1.01,0.14043840408325195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,8,1,power_law_1.01,0.1690662384033203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,8,1,power_law_1.01,0.2585536003112793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,8,1,power_law_1.01,0.27465343475341797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,8,1,power_law_1.01,2.758097839355469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,8,1,power_law_1.01,0.28204927444458006
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,8,1,power_law_1.01,0.28835391998291016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,8,1,power_law_1.01,0.29324415206909177
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,8,1,power_law_1.01,3.549501342773438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,8,1,power_law_1.01,0.30053056716918947
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,8,1,power_law_1.01,0.05027711868286133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,8,1,power_law_1.01,0.37740798950195314
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,8,1,power_law_1.01,0.39162048339843747
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,8,1,power_law_1.01,0.3903558349609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,8,1,power_law_1.01,0.39220031738281247
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,8,1,power_law_1.01,0.40506431579589847
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,8,1,power_law_1.01,0.4445049667358399
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,8,1,power_law_1.01,0.4961030578613281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,8,1,power_law_1.01,0.5540889739990235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,8,1,power_law_1.01,0.26530368804931637
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,8,1,power_law_1.01,5.260244140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,8,1,power_law_1.01,0.6514009857177735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,8,1,power_law_1.01,0.8302406311035156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,8,1,power_law_1.2,0.06932096004486084
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,8,1,power_law_1.01,1.0215122985839844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,8,1,power_law_1.2,0.0864083194732666
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,8,1,power_law_1.2,0.11881600379943849
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,8,1,power_law_1.2,0.163175048828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,8,1,power_law_1.01,1.414847412109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,8,1,power_law_1.2,0.2264531135559082
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,8,1,power_law_1.2,0.39324672698974605
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,8,1,power_law_1.01,6.7409619140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,8,1,power_law_1.2,0.40716926574707035
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,8,1,power_law_1.2,0.4303609466552734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,8,1,power_law_1.2,0.44724288940429685
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,8,1,power_law_1.2,0.46561920166015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,8,1,power_law_1.2,0.5288665771484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,8,1,power_law_1.2,0.5338035202026367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,8,1,power_law_1.01,2.5855117797851563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,8,1,power_law_1.2,0.699902114868164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,8,1,power_law_1.2,0.7115545654296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,8,1,power_law_1.01,1.8124327087402343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,8,1,power_law_1.2,0.7388518524169922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,8,1,power_law_1.2,0.7536524963378907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,8,1,power_law_1.2,0.7887789154052735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,8,1,power_law_1.01,3.3731744384765627
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,8,1,power_law_1.2,0.8706495666503906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,8,1,power_law_1.2,0.9147859191894531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,8,1,power_law_1.2,1.0649702453613281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,8,1,power_law_1.2,1.2258873748779298
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,8,1,power_law_1.2,0.05138815879821777
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,8,1,power_law_1.2,0.06328192234039307
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,8,1,power_law_1.2,0.0871008014678955
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,8,1,power_law_1.2,1.597686462402344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,8,1,power_law_1.2,0.11672191619873047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,8,1,power_law_1.2,0.16097408294677734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,8,1,power_law_1.2,0.2535807991027832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,8,1,power_law_1.2,2.0034086608886716
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,8,1,power_law_1.2,0.2642092704772949
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,8,1,power_law_1.2,0.27719999313354493
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,8,1,power_law_1.2,0.28564800262451173
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,8,1,power_law_1.2,0.2965427207946777
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,8,1,power_law_1.2,0.301023998260498
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,8,1,power_law_1.2,0.3055167961120605
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,8,1,power_law_1.2,0.3855609512329102
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,8,1,power_law_1.2,2.7680224609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,8,1,power_law_1.2,0.3902239990234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,8,1,power_law_1.2,0.3993471908569336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,8,1,power_law_1.2,0.40611968994140624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,8,1,power_law_1.2,0.42032001495361326
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,8,1,power_law_1.2,0.47318401336669924
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,8,1,power_law_1.2,0.5032287979125977
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,8,1,power_law_1.2,0.5935820770263672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,8,1,power_law_1.2,0.6545721435546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,8,1,power_law_1.2,3.571881103515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,8,1,power_law_1.2,0.8595667266845703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,1,8,balanced,0.07082496166229249
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,8,1,power_law_1.2,1.041235809326172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,1,8,balanced,0.08169088363647461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,1,8,balanced,0.11237248420715332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,1,8,balanced,0.2033932876586914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,8,1,power_law_1.2,1.422051239013672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,1,8,balanced,0.42758399963378907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,1,8,balanced,0.7121305847167969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,1,8,balanced,1.017480926513672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,8,1,power_law_1.2,1.82857666015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,1,8,balanced,1.5318156433105468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,1,8,balanced,1.5388262939453123
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,1,8,balanced,1.541327362060547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,1,8,balanced,1.5394752502441407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,1,8,balanced,1.5475340270996094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,1,8,balanced,1.5518348693847657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,1,8,balanced,1.5697485351562501
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,1,8,balanced,1.5788134765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,1,8,balanced,1.590513916015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,8,1,power_law_1.2,5.118195190429687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,1,8,balanced,1.6232012939453124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,1,8,balanced,1.6849331665039062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,1,8,balanced,1.760870361328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,1,8,balanced,1.8938124084472654
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,1,8,balanced,2.0191027832031248
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,8,1,power_law_1.2,2.65551025390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,1,8,balanced,2.2000921630859374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,1,8,balanced,2.531555786132812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,1,8,balanced,0.046449918746948246
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,1,8,balanced,0.05616640090942383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,1,8,balanced,2.8992691040039062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,1,8,balanced,0.07423871994018555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,1,8,balanced,0.12088704109191895
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,1,8,balanced,0.19864320755004883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,1,8,balanced,0.3584640121459961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,1,8,balanced,3.9351425170898438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,1,8,balanced,0.5231923294067383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,1,8,balanced,0.5588275146484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,1,8,balanced,0.5630335998535156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,1,8,balanced,0.5665536117553711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,1,8,balanced,0.5588940811157227
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,1,8,balanced,0.560714225769043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,1,8,balanced,0.5657638549804688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,1,8,balanced,0.5730419158935547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,1,8,balanced,5.420206298828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,1,8,balanced,0.582795524597168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,1,8,balanced,0.5897715377807617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,1,8,balanced,0.612353286743164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,1,8,balanced,0.6482150268554687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,1,8,balanced,0.6784844970703124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,1,8,balanced,0.7361856079101562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,1,8,balanced,0.7988172912597656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,1,8,balanced,0.9235826873779297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,1,8,balanced,1.052574691772461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,1,8,power_law_1.01,0.20346879959106445
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,1,8,balanced,7.938351440429687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,1,8,balanced,1.3002534484863282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,1,8,balanced,1.759113006591797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,1,8,power_law_1.01,0.42451198577880855
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,8,1,power_law_1.2,6.666245727539062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,1,8,power_law_1.01,0.6347980880737305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,1,8,power_law_1.01,0.7002265930175782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,1,8,balanced,2.2211929321289063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,1,8,power_law_1.01,0.7424845123291016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,1,8,power_law_1.01,0.8940505981445312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,1,8,power_law_1.01,0.860126724243164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,1,8,balanced,2.9928192138671874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,1,8,power_law_1.01,1.4509951782226562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,1,8,power_law_1.01,1.4127615356445313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,1,8,power_law_1.01,1.2722048187255859
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,8,1,power_law_1.2,3.367079772949219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,1,8,power_law_1.01,1.4432716369628906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,1,8,power_law_1.01,1.416810302734375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,1,8,power_law_1.01,1.2510873413085937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,1,8,power_law_1.01,1.366255340576172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,1,8,power_law_1.01,1.4303756713867188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,1,8,power_law_1.01,1.42582275390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,1,8,power_law_1.01,1.5011532592773438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,1,8,power_law_1.01,1.5592294311523438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,1,8,power_law_1.01,1.7765055847167968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,1,8,power_law_1.01,1.8582566833496095
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,1,8,power_law_1.01,2.1665267944335938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,1,8,power_law_1.01,0.11878911972045898
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,1,8,power_law_1.01,0.1930227279663086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,1,8,power_law_1.01,2.5212646484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,1,8,power_law_1.01,0.35034751892089844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,1,8,power_law_1.01,0.3207961654663086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,1,8,power_law_1.01,0.42174591064453126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,1,8,power_law_1.01,0.44425857543945313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,1,8,power_law_1.01,3.1843276977539063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,1,8,power_law_1.01,0.4040307235717774
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,1,8,power_law_1.01,0.5349427032470703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,1,8,power_law_1.01,0.46387966156005855
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,1,8,power_law_1.01,4.017274780273437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,1,8,power_law_1.01,0.5223475265502929
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,1,8,power_law_1.01,0.4864076614379883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,1,8,power_law_1.01,0.494890251159668
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,1,8,power_law_1.01,0.4312204742431641
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,1,8,power_law_1.01,0.5050764846801757
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,1,8,power_law_1.01,0.5014502334594726
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,1,8,power_law_1.01,0.49669631958007815
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,1,8,power_law_1.01,5.078923645019531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,1,8,power_law_1.01,0.5618086242675782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,1,8,power_law_1.01,0.5933824157714843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,1,8,power_law_1.01,0.606671371459961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,1,8,power_law_1.01,0.6755942535400391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,1,8,power_law_1.01,0.7490137481689454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,1,8,power_law_1.01,0.9122496032714844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,1,8,power_law_1.01,1.051841278076172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,1,8,power_law_1.2,0.20603519439697265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,1,8,power_law_1.01,1.4100198364257812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,1,8,power_law_1.01,7.121968383789063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,1,8,power_law_1.2,0.42816768646240233
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,1,8,power_law_1.2,0.5745203018188476
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,1,8,power_law_1.01,1.8395968627929686
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,1,8,power_law_1.2,0.6460428619384766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,1,8,power_law_1.2,0.8100057220458984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,1,8,power_law_1.01,2.474705963134766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,1,8,power_law_1.2,0.9337318420410157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,1,8,power_law_1.2,1.0089318084716798
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,1,8,power_law_1.2,1.3106866455078126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,1,8,power_law_1.01,3.479344787597656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,1,8,power_law_1.2,1.4038079833984374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,1,8,power_law_1.01,9.4216064453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,1,8,power_law_1.2,1.2889637756347656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,1,8,power_law_1.2,1.3439642333984376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,1,8,power_law_1.2,1.3045747375488281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,1,8,power_law_1.2,1.3528909301757812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,1,8,power_law_1.2,1.4175129699707032
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,1,8,power_law_1.2,1.3590911865234374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,1,8,power_law_1.2,1.4330995178222656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,1,8,power_law_1.2,1.5105702209472658
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,1,8,power_law_1.2,1.7267738342285157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,1,8,power_law_1.2,1.7571775817871091
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,1,8,power_law_1.2,1.940130615234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,1,8,power_law_1.2,0.11885951995849608
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,1,8,power_law_1.2,2.331676177978516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,1,8,power_law_1.2,0.19442687988281251
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,1,8,power_law_1.2,2.7496551513671874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,1,8,power_law_1.2,0.2935206413269043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,1,8,power_law_1.2,0.3221900939941406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,1,8,power_law_1.2,3.3667276000976565
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,1,8,power_law_1.2,0.420634880065918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,1,8,power_law_1.2,0.47528705596923826
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,1,8,power_law_1.2,0.46140289306640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,1,8,power_law_1.2,0.5262745666503906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,1,8,power_law_1.2,0.4542591857910156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,1,8,power_law_1.2,0.4664499282836914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,1,8,power_law_1.2,4.718023681640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,1,8,power_law_1.2,0.47661567687988277
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,1,8,power_law_1.2,0.46061054229736326
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,1,8,power_law_1.2,0.4660748672485352
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,1,8,power_law_1.2,0.47815296173095706
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,1,8,power_law_1.2,0.48422401428222656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,1,8,power_law_1.2,0.5511334228515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,1,8,power_law_1.2,0.5973990249633789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,1,8,power_law_1.2,5.860173950195312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,1,8,power_law_1.2,0.6266393661499023
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,1,8,power_law_1.2,0.7065740966796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,1,8,power_law_1.2,0.5022604751586914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,1,8,power_law_1.2,0.8032179260253907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,1,8,power_law_1.2,0.9389862060546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,2,4,balanced,0.06895872116088866
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,1,8,power_law_1.2,1.2082355499267579
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,2,4,balanced,0.08080512046813965
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,1,8,power_law_1.2,7.560748901367188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,2,4,balanced,0.11791999816894531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,1,8,power_law_1.2,1.6255871582031252
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,2,4,balanced,0.1895193672180176
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,2,4,balanced,0.41968128204345706
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,2,4,balanced,0.6969497680664063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,1,8,power_law_1.2,1.9817958068847656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,2,4,balanced,1.0085734558105468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,2,4,balanced,1.0115187072753906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,2,4,balanced,1.033354263305664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,2,4,balanced,1.02063232421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,1,8,power_law_1.2,2.9327642822265623
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,2,4,balanced,1.5003584289550782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,2,4,balanced,1.5059046936035156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,2,4,balanced,1.5144677734375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,2,4,balanced,1.5280563354492187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,2,4,balanced,1.5367411804199218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,2,4,balanced,1.5492250061035155
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,1,8,power_law_1.2,11.225687255859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,1,8,power_law_1.2,3.743210144042969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,2,4,balanced,1.5831167602539062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,2,4,balanced,1.631392059326172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,2,4,balanced,1.699957733154297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,2,4,balanced,1.8531826782226564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,2,4,balanced,0.04141568183898926
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,2,4,balanced,1.9847975158691404
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,2,4,balanced,0.0530291223526001
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,2,4,balanced,0.0717248010635376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,2,4,balanced,2.1717312622070315
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,2,4,balanced,0.11642496109008789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,2,4,balanced,0.19380992889404297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,2,4,balanced,0.35949695587158204
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,2,4,balanced,2.513799743652344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,2,4,balanced,0.5169625473022461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,2,4,balanced,0.5201049423217773
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,2,4,balanced,0.5224089431762695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,2,4,balanced,0.5272447967529297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,2,4,balanced,0.563823356628418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,2,4,balanced,0.56611328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,2,4,balanced,0.5695180892944336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,2,4,balanced,3.84466552734375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,2,4,balanced,0.5795724868774415
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,2,4,balanced,2.865186462402344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,2,4,balanced,0.5858329772949219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,2,4,balanced,0.5909516906738281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,2,4,balanced,0.6140326309204102
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,2,4,balanced,0.6461363220214844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,2,4,balanced,0.6645862579345703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,2,4,balanced,0.7184063720703124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,2,4,balanced,0.7840857696533203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,2,4,balanced,5.172243041992187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,2,4,balanced,0.8766067504882813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,2,4,power_law_1.01,0.11626751899719238
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,2,4,balanced,0.9997747039794922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,2,4,power_law_1.01,0.1842470359802246
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,2,4,balanced,1.2210201263427733
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,2,4,power_law_1.01,0.40782081604003906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,2,4,power_law_1.01,0.4131008148193359
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,2,4,balanced,1.7363352966308594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,2,4,power_law_1.01,0.5485657501220703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,2,4,power_law_1.01,0.6227468872070313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,2,4,balanced,2.112728271484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,2,4,balanced,7.270352783203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,2,4,power_law_1.01,0.7427059173583984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,2,4,power_law_1.01,0.8616614532470702
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,2,4,power_law_1.01,0.8625536346435547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,2,4,power_law_1.01,0.8975872039794922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,2,4,balanced,2.8784307861328124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,2,4,power_law_1.01,1.3260389709472657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,2,4,power_law_1.01,1.335800323486328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,2,4,power_law_1.01,1.3419532775878906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,2,4,power_law_1.01,1.311082305908203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,2,4,power_law_1.01,1.3213247680664062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,2,4,power_law_1.01,1.3499813842773438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,2,4,power_law_1.01,1.4373274230957032
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,2,4,power_law_1.01,1.4892300415039064
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,2,4,power_law_1.01,1.552042236328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,2,4,power_law_1.01,1.7248320007324218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,2,4,power_law_1.01,1.9272140502929687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,2,4,power_law_1.01,0.0718067216873169
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,2,4,power_law_1.01,0.11956607818603515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,2,4,power_law_1.01,2.3554202270507814
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,2,4,power_law_1.01,0.19283327102661132
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,2,4,power_law_1.01,2.6857522583007816
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,2,4,power_law_1.01,0.20209024429321287
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,2,4,power_law_1.01,0.2812428855895996
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,2,4,power_law_1.01,0.3147865676879883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,2,4,power_law_1.01,0.39057537078857424
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,2,4,power_law_1.01,0.4181631851196289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,2,4,power_law_1.01,0.4427955245971679
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,2,4,power_law_1.01,0.4262809753417969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,2,4,power_law_1.01,3.59581298828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,2,4,power_law_1.01,0.48163200378417964
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,2,4,power_law_1.01,0.4956659317016602
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,2,4,power_law_1.01,0.5078028869628907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,2,4,power_law_1.01,0.49678974151611327
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,2,4,power_law_1.01,0.4912255859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,2,4,power_law_1.01,0.5329075241088868
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,2,4,power_law_1.01,4.4191308593750005
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,2,4,power_law_1.01,0.5394432067871093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,2,4,power_law_1.01,0.5545011138916015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,2,4,power_law_1.01,0.6067161560058594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,2,4,power_law_1.01,0.6677875518798828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,2,4,power_law_1.01,0.7613056182861329
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,2,4,power_law_1.01,0.9081024169921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,2,4,power_law_1.01,1.029186553955078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,2,4,power_law_1.2,0.11670656204223633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,2,4,power_law_1.01,6.189013671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,2,4,power_law_1.2,0.18373504638671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,2,4,power_law_1.01,1.3994636535644531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,2,4,power_law_1.2,0.32856063842773436
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,2,4,power_law_1.2,0.4227596664428711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,2,4,power_law_1.01,1.781538543701172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,2,4,power_law_1.2,0.5158259201049804
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,2,4,power_law_1.2,0.6256358337402343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,2,4,power_law_1.2,0.7356018829345703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,2,4,power_law_1.01,2.480432586669922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,2,4,power_law_1.01,7.666184692382812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,2,4,power_law_1.2,0.8741593933105468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,2,4,power_law_1.2,0.8467404937744141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,2,4,power_law_1.2,0.9047936248779298
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,2,4,power_law_1.01,3.069219970703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,2,4,power_law_1.2,1.3835877990722656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,2,4,power_law_1.2,1.2759423828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,2,4,power_law_1.2,1.2818841552734375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,2,4,power_law_1.2,1.236948471069336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,2,4,power_law_1.2,1.3229656982421876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,2,4,power_law_1.2,1.3955711364746093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,2,4,power_law_1.2,1.3886297607421876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,2,4,power_law_1.2,1.5285504150390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,2,4,power_law_1.2,1.6225523376464843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,2,4,power_law_1.2,1.8149478149414062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,2,4,power_law_1.2,0.07136767864227295
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,2,4,power_law_1.2,2.061397705078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,2,4,power_law_1.2,0.11517824172973631
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,2,4,power_law_1.2,2.546534423828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,2,4,power_law_1.2,0.1678835105895996
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,2,4,power_law_1.2,2.8510861206054687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,2,4,power_law_1.2,0.20153600692749024
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,2,4,power_law_1.2,0.2747776031494141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,2,4,power_law_1.2,0.31503488540649416
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,2,4,power_law_1.2,0.383430404663086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,2,4,power_law_1.2,0.44620288848876954
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,2,4,power_law_1.2,0.43632766723632815
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,2,4,power_law_1.2,0.4135500717163086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,2,4,power_law_1.2,3.992948608398437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,2,4,power_law_1.2,0.46307456970214844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,2,4,power_law_1.2,0.4659686279296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,2,4,power_law_1.2,0.48554752349853514
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,2,4,power_law_1.2,0.47593215942382816
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,2,4,power_law_1.2,4.7510067749023435
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,2,4,power_law_1.2,0.4903769683837891
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,2,4,power_law_1.2,0.4955942535400391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,2,4,power_law_1.2,0.5159603118896484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,2,4,power_law_1.2,0.5764416122436524
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,2,4,power_law_1.2,0.6095219039916993
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,2,4,power_law_1.2,0.7025587463378906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,2,4,power_law_1.2,0.7788761901855469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,2,4,power_law_1.2,0.9241766357421876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,2,4,power_law_1.2,1.1124928283691407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,4,2,balanced,0.06407423973083495
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,4,2,balanced,0.10863360404968261
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,4,2,balanced,0.07902336120605469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,2,4,power_law_1.2,6.568914184570312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,4,2,balanced,0.18164480209350586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,2,4,power_law_1.2,1.5700057983398437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,4,2,balanced,0.4500492858886719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,4,2,balanced,0.6950605010986328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,4,2,balanced,1.0022259521484376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,2,4,power_law_1.2,2.7032513427734375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,4,2,balanced,1.0080025482177735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,2,4,power_law_1.2,1.8828172302246096
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,4,2,balanced,1.0099225616455079
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,2,4,power_law_1.2,3.4298803710937498
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,4,2,balanced,1.015906524658203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,4,2,balanced,1.0249394989013672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,2,4,power_law_1.2,9.281390380859374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,4,2,balanced,1.0302003479003905
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,4,2,balanced,1.0377407836914063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,4,2,balanced,1.499764404296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,4,2,balanced,1.514086456298828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,4,2,balanced,1.5270796203613282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,4,2,balanced,1.5629592895507813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,4,2,balanced,1.605048370361328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,4,2,balanced,1.668695068359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,4,2,balanced,1.797137908935547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,4,2,balanced,1.9333261108398438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,4,2,balanced,0.0414079999923706
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,4,2,balanced,0.054807038307189936
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,4,2,balanced,0.07529088020324706
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,4,2,balanced,0.12348031997680664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,4,2,balanced,2.157637176513672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,4,2,balanced,0.20527103424072263
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,4,2,balanced,0.381841926574707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,4,2,balanced,2.6112115478515627
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,4,2,balanced,0.5539558410644532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,4,2,balanced,2.9132415771484377
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,4,2,balanced,0.5605542373657226
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,4,2,balanced,0.5648204803466796
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,4,2,balanced,0.5648307037353516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,4,2,balanced,0.5761446380615235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,4,2,balanced,0.5518207931518555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,4,2,balanced,3.8480090332031254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,4,2,balanced,0.6454630279541016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,4,2,balanced,0.6489395141601563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,4,2,balanced,0.6559308624267578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,4,2,balanced,0.6659315490722657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,4,2,balanced,0.7020134735107422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,4,2,balanced,0.7201792144775391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,4,2,balanced,0.5775743865966797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,4,2,balanced,0.7653247833251953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,4,2,balanced,5.185270996093751
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,4,2,balanced,0.8205785369873047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,4,2,balanced,0.9182796478271484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,4,2,power_law_1.01,0.07970047950744628
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,4,2,balanced,1.0442022705078124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,4,2,power_law_1.01,0.10873855590820311
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,4,2,balanced,1.2364275360107422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,4,2,power_law_1.01,0.1782476806640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,4,2,power_law_1.01,0.24772480010986325
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,4,2,power_law_1.01,0.42001792907714847
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,4,2,balanced,1.7789773559570314
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,4,2,power_law_1.01,0.5006886291503906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,4,2,power_law_1.01,0.626437110900879
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,4,2,balanced,2.154407653808594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,4,2,power_law_1.01,0.7439577484130859
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,4,2,balanced,7.20919189453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,4,2,power_law_1.01,0.7396147155761719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,4,2,power_law_1.01,0.8046707153320313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,4,2,power_law_1.01,0.8083340454101563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,4,2,power_law_1.01,0.8903040313720704
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,4,2,balanced,2.94260986328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,4,2,power_law_1.01,0.901312026977539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,4,2,power_law_1.01,1.3070938110351562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,4,2,power_law_1.01,1.3151718139648438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,4,2,power_law_1.01,1.3528448486328126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,4,2,power_law_1.01,1.3680345153808593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,4,2,power_law_1.01,1.4905932617187498
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,4,2,power_law_1.01,1.55931640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,4,2,power_law_1.01,1.723047637939453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,4,2,power_law_1.01,1.945296630859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,4,2,power_law_1.01,0.05502848148345947
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,4,2,power_law_1.01,0.07513728141784667
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,4,2,power_law_1.01,2.2518182373046876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,4,2,power_law_1.01,0.122741756439209
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,4,2,power_law_1.01,0.15614336013793945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,4,2,power_law_1.01,2.65749755859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,4,2,power_law_1.01,0.23121280670166017
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,4,2,power_law_1.01,0.2853683280944824
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,4,2,power_law_1.01,0.32928512573242186
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,4,2,power_law_1.01,3.3115457153320316
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,4,2,power_law_1.01,0.42787967681884764
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,4,2,power_law_1.01,0.4457727813720703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,4,2,power_law_1.01,0.46617855072021486
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,4,2,power_law_1.01,0.48781696319580076
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,4,2,power_law_1.01,0.4910195159912109
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,4,2,power_law_1.01,4.0522790527343755
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,4,2,power_law_1.01,0.544692497253418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,4,2,power_law_1.01,0.42534015655517576
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,4,2,power_law_1.01,0.524853744506836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,4,2,power_law_1.01,0.537578239440918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,4,2,power_law_1.01,0.5843737411499024
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,4,2,power_law_1.01,0.6405709075927735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,4,2,power_law_1.01,0.6482367706298828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,4,2,power_law_1.01,0.7178201293945312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,4,2,power_law_1.01,0.7932608032226562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,4,2,power_law_1.01,0.9593536376953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,4,2,power_law_1.01,5.681489868164062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,4,2,power_law_1.2,0.08062080383300782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,4,2,power_law_1.01,1.1383360290527345
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,4,2,power_law_1.2,0.10861056327819825
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,4,2,power_law_1.01,1.4431079101562498
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,4,2,power_law_1.2,0.15403648376464846
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,4,2,power_law_1.01,1.796136932373047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,4,2,power_law_1.2,0.4120064163208008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,4,2,power_law_1.2,0.489728012084961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,4,2,power_law_1.01,7.315693969726563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,4,2,power_law_1.2,0.5179647827148438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,4,2,power_law_1.01,2.4763827514648438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,4,2,power_law_1.2,0.24261760711669922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,4,2,power_law_1.2,0.7302207946777344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,4,2,power_law_1.2,0.7535923004150391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,4,2,power_law_1.2,0.7647705841064453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,4,2,power_law_1.2,0.7818406677246094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,4,2,power_law_1.01,3.0982733154296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,4,2,power_law_1.2,0.8645516967773437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,4,2,power_law_1.2,1.2967706298828126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,4,2,power_law_1.2,1.3249842834472656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,4,2,power_law_1.2,1.3769728088378907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,4,2,power_law_1.2,1.3392410278320312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,4,2,power_law_1.2,0.8476044464111329
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,4,2,power_law_1.2,1.5215449523925781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,4,2,power_law_1.2,1.6192576599121093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,4,2,power_law_1.2,1.7873266601562499
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,4,2,power_law_1.2,0.05458687782287598
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,4,2,power_law_1.2,0.07530496120452881
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,4,2,power_law_1.2,2.008257293701172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,4,2,power_law_1.2,2.308771820068359
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,4,2,power_law_1.2,0.10696448326110838
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,4,2,power_law_1.2,0.1578444766998291
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,4,2,power_law_1.2,2.764801330566406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,4,2,power_law_1.2,0.2180928039550781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,4,2,power_law_1.2,0.2627699279785156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,4,2,power_law_1.2,0.29085311889648435
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,4,2,power_law_1.2,3.4227685546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,4,2,power_law_1.2,0.41269889831542966
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,4,2,power_law_1.2,0.4236928176879883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,4,2,power_law_1.2,0.41727745056152343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,4,2,power_law_1.2,0.4330137634277344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,4,2,power_law_1.2,0.4624230575561524
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,4,2,power_law_1.2,0.4739712142944336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,4,2,power_law_1.2,0.5523468780517578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,4,2,power_law_1.2,0.520761604309082
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,4,2,power_law_1.2,0.5365785598754883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,4,2,power_law_1.2,0.5872819137573242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,4,2,power_law_1.2,0.633831672668457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,4,2,power_law_1.2,0.6695091247558593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,4,2,power_law_1.2,4.198426818847656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,4,2,power_law_1.2,0.7445849609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,4,2,power_law_1.2,0.8320230102539062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,4,2,power_law_1.2,5.858512573242187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,4,2,power_law_1.2,1.0068045043945313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,8,1,balanced,0.07915840148925782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,4,2,power_law_1.2,1.1869696044921876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,8,1,balanced,0.08995264053344727
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,8,1,balanced,0.12020480155944824
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,8,1,balanced,0.1916595268249512
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,8,1,balanced,0.4225260925292969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,4,2,power_law_1.2,1.82600830078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,8,1,balanced,0.7080313873291015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,4,2,power_law_1.2,7.758905639648438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,4,2,power_law_1.2,2.5673703002929686
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,4,2,power_law_1.2,1.5169842529296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,8,1,balanced,1.0275257873535157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,8,1,balanced,1.0288070678710937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,8,1,balanced,1.0339609527587892
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,4,2,power_law_1.2,3.273483581542969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,8,1,balanced,1.035846405029297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,8,1,balanced,1.0470687866210937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,8,1,balanced,1.051981430053711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,8,1,balanced,1.0611167907714845
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,8,1,balanced,1.08508544921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,8,1,balanced,1.0944665527343749
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,8,1,balanced,1.109553909301758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,8,1,balanced,1.5373490905761718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,8,1,balanced,1.5804576110839843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,8,1,balanced,1.6734124755859376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,8,1,balanced,1.7779481506347659
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,8,1,balanced,1.9208198547363282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,8,1,balanced,0.05412479877471924
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,8,1,balanced,0.07017024040222168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,8,1,balanced,2.190229187011719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,8,1,balanced,0.09169728279113769
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,8,1,balanced,0.14495807647705078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,8,1,balanced,0.2495008087158203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,8,1,balanced,0.4620755386352539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,8,1,balanced,0.6705497741699219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,8,1,balanced,3.0742642211914064
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,8,1,balanced,0.6725901031494141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,8,1,balanced,0.6755123138427734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,8,1,balanced,0.6772096252441406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,8,1,balanced,0.6826911926269531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,8,1,balanced,0.6861395263671876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,8,1,balanced,2.9720382690429688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,8,1,balanced,0.6900870513916015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,8,1,balanced,3.932185668945313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,8,1,balanced,0.6984639739990234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,8,1,balanced,0.7051455688476562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,8,1,balanced,0.7143993377685547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,8,1,balanced,0.8450323486328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,8,1,balanced,0.8777702331542969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,8,1,balanced,0.8819961547851562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,8,1,balanced,0.9230713653564454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,8,1,balanced,0.9766687774658204
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,8,1,balanced,5.3246923828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,8,1,balanced,1.0776416015624999
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,8,1,power_law_1.01,0.0752127981185913
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,8,1,balanced,1.1886450958251953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,8,1,power_law_1.01,0.08849408149719237
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,8,1,balanced,1.40400390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,8,1,power_law_1.01,0.11943424224853516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,8,1,power_law_1.01,0.19056447982788086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,8,1,power_law_1.01,0.40996479034423833
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,8,1,power_law_1.01,0.4958534240722656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,8,1,balanced,2.1062136840820314
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,8,1,power_law_1.01,0.5649375915527344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,8,1,balanced,7.5216217041015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,8,1,power_law_1.01,0.7224313354492187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,8,1,power_law_1.01,0.7529599761962891
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,8,1,power_law_1.01,0.7813638305664062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,8,1,power_law_1.01,0.8087686157226562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,8,1,power_law_1.01,0.8210745239257813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,8,1,balanced,2.494051818847656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,8,1,balanced,3.4322680664062504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,8,1,power_law_1.01,0.8814701080322266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,8,1,power_law_1.01,0.9229606628417969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,8,1,power_law_1.01,0.9434496307373047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,8,1,power_law_1.01,1.0194009399414061
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,8,1,power_law_1.01,1.3475564575195311
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,8,1,power_law_1.01,1.4094522094726563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,8,1,power_law_1.01,1.5154112243652345
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,8,1,power_law_1.01,1.723014373779297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,8,1,power_law_1.01,1.8879168701171873
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,8,1,power_law_1.01,0.05458879947662354
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,8,1,power_law_1.01,2.2697772216796874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,8,1,power_law_1.01,0.06985472202301025
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,8,1,power_law_1.01,0.089749116897583
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,8,1,power_law_1.01,0.14333824157714842
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,8,1,power_law_1.01,2.648029479980469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,8,1,power_law_1.01,0.23368255615234376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,8,1,power_law_1.01,0.3226118469238281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,8,1,power_law_1.01,0.3751808166503906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,8,1,power_law_1.01,0.4676371383666992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,8,1,power_law_1.01,0.491064338684082
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,8,1,power_law_1.01,3.294300231933594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,8,1,power_law_1.01,0.5159596633911133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,8,1,power_law_1.01,0.5339385604858398
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,8,1,power_law_1.01,0.5409920120239258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,8,1,power_law_1.01,0.5556607818603515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,8,1,power_law_1.01,0.5753843307495117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,8,1,power_law_1.01,4.025521240234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,8,1,power_law_1.01,0.5804249572753906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,8,1,power_law_1.01,0.6017184066772461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,8,1,power_law_1.01,0.717237777709961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,8,1,power_law_1.01,0.7430220794677734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,8,1,power_law_1.01,0.8034400177001952
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,8,1,power_law_1.01,0.8788793945312501
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,8,1,power_law_1.01,0.9522988891601563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,8,1,power_law_1.01,5.58336669921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,8,1,power_law_1.01,1.1144134521484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,8,1,power_law_1.2,0.07876800060272217
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,8,1,power_law_1.01,1.294493408203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,8,1,power_law_1.2,0.08879039764404298
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,8,1,power_law_1.2,0.12102848052978517
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,8,1,power_law_1.01,1.6507583618164063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,8,1,power_law_1.2,0.1909894371032715
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,8,1,power_law_1.2,0.37229118347167967
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,8,1,power_law_1.01,2.0282252502441405
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,8,1,power_law_1.2,0.4622028732299805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,8,1,power_law_1.2,0.5294777679443359
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,8,1,power_law_1.01,7.320741577148437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,8,1,power_law_1.2,0.7109414672851562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,8,1,power_law_1.2,0.7361235046386719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,8,1,power_law_1.01,2.785868835449219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,8,1,power_law_1.2,0.7580786895751953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,8,1,power_law_1.2,0.7861119842529296
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,8,1,power_law_1.2,0.8269010925292969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,8,1,power_law_1.2,0.9020563507080078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,8,1,power_law_1.2,0.9161996459960937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,8,1,power_law_1.01,3.5381304931640627
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,8,1,power_law_1.2,0.9600473785400391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,8,1,power_law_1.2,1.0213657379150392
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,8,1,power_law_1.2,1.3445292663574218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,8,1,power_law_1.2,1.4369715881347656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,8,1,power_law_1.2,1.5508569335937499
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,8,1,power_law_1.2,1.7589414978027342
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,8,1,power_law_1.2,1.9553382873535157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,8,1,power_law_1.2,0.05425216197967529
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,8,1,power_law_1.2,0.06948095798492432
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,8,1,power_law_1.2,0.09131135940551757
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,8,1,power_law_1.2,2.3362841796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,8,1,power_law_1.2,2.676147155761719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,8,1,power_law_1.2,0.1431769561767578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,8,1,power_law_1.2,0.21551231384277342
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,8,1,power_law_1.2,0.30310592651367185
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,8,1,power_law_1.2,0.3473561477661133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,8,1,power_law_1.2,0.4592326354980469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,8,1,power_law_1.2,0.4782195281982422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,8,1,power_law_1.2,0.49603904724121095
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,8,1,power_law_1.2,3.364676513671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,8,1,power_law_1.2,4.111587219238281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,8,1,power_law_1.2,0.5151968002319336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,8,1,power_law_1.2,0.5382054519653321
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,8,1,power_law_1.2,0.5501100921630859
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,8,1,power_law_1.2,0.5681292724609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,8,1,power_law_1.2,0.58451904296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,8,1,power_law_1.2,0.5956134414672851
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,8,1,power_law_1.2,0.715428466796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,8,1,power_law_1.2,0.7790406036376953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,8,1,power_law_1.2,5.631294555664063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,8,1,power_law_1.2,0.8176710510253906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,8,1,power_law_1.2,0.9047641754150391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,8,1,power_law_1.2,1.0026624298095703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,1,8,balanced,0.025121281147003172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,1,8,balanced,0.0438489580154419
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,8,1,power_law_1.2,1.184958724975586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,1,8,balanced,0.051869440078735354
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,1,8,balanced,0.07545472145080566
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,1,8,balanced,0.1288102436065674
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,8,1,power_law_1.2,1.36611328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,1,8,balanced,0.23387392044067382
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,1,8,balanced,0.34733440399169924
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,1,8,balanced,0.35318145751953123
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,1,8,balanced,0.3623936080932617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,8,1,power_law_1.2,1.718551025390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,1,8,balanced,0.36752254486083985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,8,1,power_law_1.2,7.230829467773438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,1,8,balanced,0.3596543884277344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,1,8,balanced,0.35995136260986327
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,1,8,balanced,0.3612953567504883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,1,8,balanced,0.368798713684082
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,1,8,balanced,0.3887756729125976
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,1,8,balanced,0.38517505645751954
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,1,8,balanced,0.40302974700927735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,1,8,balanced,0.43299198150634766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,1,8,balanced,0.44916862487792975
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,1,8,balanced,0.4890291213989258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,1,8,balanced,0.5193638229370118
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,1,8,balanced,0.6353587341308594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,1,8,balanced,0.7299366760253906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,8,1,power_law_1.2,2.8168679809570314
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,1,8,balanced,0.026878719329833982
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,1,8,balanced,1.0271475219726562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,1,8,balanced,0.028625919818878177
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,1,8,balanced,0.03600895881652832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,1,8,balanced,0.051210241317749025
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,1,8,balanced,0.08383872032165526
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,1,8,balanced,1.3053234863281251
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,1,8,balanced,0.15463040351867677
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,1,8,balanced,0.15786879539489745
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,1,8,balanced,0.16060287475585938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,1,8,balanced,0.1619481658935547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,1,8,balanced,0.15468544006347656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,8,1,power_law_1.2,3.6129464721679687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,1,8,balanced,0.15583104133605957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,1,8,balanced,0.1566540813446045
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,1,8,balanced,1.9202648925781252
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,1,8,balanced,0.15981823921203614
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,1,8,balanced,0.1665817642211914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,1,8,balanced,0.17223936080932617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,1,8,balanced,0.1826278305053711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,1,8,balanced,0.21820543289184569
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,1,8,balanced,0.2544921684265137
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,1,8,balanced,0.2849983978271484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,1,8,balanced,0.1370150375366211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,1,8,balanced,0.3553420639038086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,1,8,balanced,2.5049395751953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,1,8,balanced,0.42756607055664064
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,1,8,power_law_1.01,0.07180416107177734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,1,8,balanced,0.5994419097900391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,8,1,power_law_1.2,2.1495680236816406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,1,8,balanced,0.7336358642578125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,1,8,power_law_1.01,0.1734886360168457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,1,8,balanced,0.20540416717529295
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,1,8,power_law_1.01,0.19699071884155273
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,1,8,power_law_1.01,0.19266687393188478
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,1,8,balanced,1.0565580749511718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,1,8,power_law_1.01,0.19647872924804688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,1,8,power_law_1.01,0.30534015655517577
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,1,8,power_law_1.01,0.32629249572753904
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,1,8,power_law_1.01,0.3189606475830078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,1,8,balanced,1.3818521118164062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,1,8,power_law_1.01,0.33136127471923826
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,1,8,power_law_1.01,0.3538457489013672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,1,8,power_law_1.01,0.3466099166870117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,1,8,power_law_1.01,0.12361599922180175
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,1,8,power_law_1.01,0.3572800064086914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,1,8,power_law_1.01,0.39053951263427733
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,1,8,power_law_1.01,0.37696510314941406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,1,8,power_law_1.01,0.40807167053222654
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,1,8,power_law_1.01,0.44880512237548825
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,1,8,power_law_1.01,0.4984972763061523
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,1,8,power_law_1.01,0.5623129653930664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,1,8,power_law_1.01,0.7119705963134766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,1,8,power_law_1.01,0.3490150451660156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,1,8,power_law_1.01,0.8622220611572265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,1,8,power_law_1.01,0.3621811294555664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,1,8,power_law_1.01,0.08136320114135742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,1,8,power_law_1.01,1.1751974487304688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,1,8,power_law_1.01,0.09330816268920898
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,1,8,power_law_1.01,0.10018943786621093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,1,8,power_law_1.01,0.11639807701110841
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,1,8,power_law_1.01,0.1129689598083496
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,1,8,power_law_1.01,0.14898943901062012
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,1,8,power_law_1.01,0.14320639610290525
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,1,8,power_law_1.01,0.1480729579925537
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,1,8,power_law_1.01,0.04874368190765381
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,1,8,power_law_1.01,0.14713343620300293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,1,8,power_law_1.01,0.15195263862609862
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,1,8,power_law_1.01,0.15125120162963868
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,1,8,power_law_1.01,0.15182592391967772
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,1,8,power_law_1.01,0.15860095977783203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,1,8,power_law_1.01,2.4027558898925783
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,1,8,power_law_1.01,0.16225151062011717
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,1,8,power_law_1.01,1.6795762634277345
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,1,8,power_law_1.01,0.24376192092895507
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,1,8,power_law_1.01,0.2776435279846191
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,1,8,power_law_1.01,0.15409024238586427
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,1,8,power_law_1.01,3.1505767822265627
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,1,8,power_law_1.01,0.34429695129394533
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,1,8,power_law_1.01,0.43277568817138673
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,1,8,power_law_1.01,0.1615577507019043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,1,8,power_law_1.2,0.07122687816619873
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,1,8,power_law_1.01,0.18663936614990234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,1,8,power_law_1.01,0.6136243057250976
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,1,8,power_law_1.2,0.12279040336608886
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,1,8,power_law_1.01,0.21191551208496096
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,1,8,power_law_1.2,0.15954303741455078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,1,8,power_law_1.2,0.19453567504882813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,1,8,power_law_1.2,0.19151872634887696
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,1,8,power_law_1.2,0.21561344146728514
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,1,8,power_law_1.01,1.116275177001953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,1,8,power_law_1.2,0.31764223098754885
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,1,8,power_law_1.2,0.3280524826049805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,1,8,power_law_1.2,0.32656639099121093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,1,8,power_law_1.2,0.35854846954345704
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,1,8,power_law_1.2,0.33504257202148435
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,1,8,power_law_1.01,0.7709696197509766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,1,8,power_law_1.2,0.35247360229492186
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,1,8,power_law_1.2,0.36199680328369144
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,1,8,power_law_1.2,0.370887680053711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,1,8,power_law_1.2,0.395098876953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,1,8,power_law_1.2,0.30678911209106446
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,1,8,power_law_1.2,0.4167155075073242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,1,8,power_law_1.2,0.455810546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,1,8,power_law_1.2,0.5359219360351563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,1,8,power_law_1.2,0.6077734375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,1,8,power_law_1.2,0.7958105468750001
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,1,8,power_law_1.2,1.0843609619140626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,1,8,power_law_1.2,0.0492633581161499
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,1,8,power_law_1.2,0.08033151626586914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,1,8,power_law_1.2,0.088024320602417
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,1,8,power_law_1.2,1.4349440002441407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,1,8,power_law_1.2,0.09730815887451172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,1,8,power_law_1.2,0.11335552215576172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,1,8,power_law_1.2,0.1230463981628418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,1,8,power_law_1.2,1.8986303710937498
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,1,8,power_law_1.2,0.14243583679199218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,1,8,power_law_1.2,0.14674176216125487
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,1,8,power_law_1.2,0.3177689552307129
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,1,8,power_law_1.2,0.14729855537414552
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,1,8,power_law_1.2,0.15762175559997557
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,1,8,power_law_1.2,0.15453184127807618
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,1,8,power_law_1.01,1.407317810058594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,1,8,power_law_1.2,0.1409011173248291
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,1,8,power_law_1.2,0.1552768039703369
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,1,8,power_law_1.2,0.15447423934936524
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,1,8,power_law_1.2,0.16151424407958986
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,1,8,power_law_1.2,0.17027456283569337
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,1,8,power_law_1.2,2.8656524658203124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,1,8,power_law_1.2,0.19531648635864257
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,1,8,power_law_1.2,0.22094720840454102
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,1,8,power_law_1.2,0.26388479232788087
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,1,8,power_law_1.2,0.30615552902221677
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,1,8,power_law_1.2,0.149936637878418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,1,8,power_law_1.2,0.39158657073974606
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,1,8,power_law_1.2,0.5048153686523438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,2,4,balanced,0.02268928050994873
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,2,4,balanced,0.03846271991729736
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,1,8,power_law_1.2,0.6641203308105469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,1,8,power_law_1.2,3.5522906494140627
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,2,4,balanced,0.04912511825561523
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,2,4,balanced,0.07317759990692138
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,1,8,power_law_1.2,0.8416422271728516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,2,4,balanced,0.2350553512573242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,2,4,balanced,0.3366579055786133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,2,4,balanced,0.33927551269531253
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,1,8,power_law_1.2,1.2804441833496094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,2,4,balanced,0.34063743591308593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,2,4,balanced,0.349040641784668
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,2,4,balanced,0.34984062194824217
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,1,8,power_law_1.2,1.6492965698242188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,2,4,balanced,0.3571673583984375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,2,4,balanced,0.37121662139892575
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,2,4,balanced,0.12385151863098146
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,2,4,balanced,0.38264831542968747
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,2,4,balanced,0.23382272720336914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,2,4,balanced,0.41257343292236326
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,2,4,balanced,0.43315071105957037
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,2,4,balanced,0.4707379150390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,2,4,balanced,0.5031091308593749
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,2,4,balanced,0.6106982421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,2,4,balanced,0.3641011047363281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,2,4,balanced,0.024528639316558836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,2,4,balanced,0.22754688262939454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,2,4,balanced,0.9688690948486329
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,2,4,balanced,0.02668800115585327
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,2,4,balanced,0.035411200523376464
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,2,4,balanced,0.05215487957000733
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,2,4,balanced,0.08255231857299805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,2,4,balanced,1.2429901123046876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,2,4,balanced,0.13695743560791016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,2,4,balanced,0.15451135635375976
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,2,4,balanced,0.7016102600097656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,2,4,balanced,0.15425919532775878
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,2,4,balanced,0.15596799850463866
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,2,4,balanced,1.801457977294922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,2,4,balanced,0.15682944297790527
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,2,4,balanced,0.15790592193603517
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,2,4,balanced,0.16173952102661132
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,2,4,balanced,0.16504447937011718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,2,4,balanced,0.16913663864135742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,2,4,balanced,0.17743104934692383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,2,4,balanced,0.1995583915710449
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,2,4,balanced,2.332919006347656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,2,4,balanced,0.20798719406127927
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,2,4,balanced,0.24029056549072267
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,2,4,balanced,0.2715135955810547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,2,4,balanced,0.1398681640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,2,4,balanced,0.32023681640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,2,4,power_law_1.01,0.04821887969970703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,2,4,balanced,0.3868608093261719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,2,4,power_law_1.01,0.06979072093963623
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,2,4,balanced,0.5526310348510742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,2,4,power_law_1.01,0.09265279769897461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,2,4,balanced,0.6632102203369141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,2,4,power_law_1.01,0.10902912139892577
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,2,4,power_law_1.01,0.15223551750183106
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,2,4,power_law_1.01,0.18107519149780274
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,2,4,power_law_1.01,0.20032255172729493
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,2,4,power_law_1.01,0.20343679428100586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,2,4,power_law_1.01,0.3051724815368652
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,2,4,balanced,0.13912063598632812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,2,4,balanced,1.235560989379883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,2,4,power_law_1.01,0.3131520080566406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,2,4,power_law_1.01,0.31559936523437504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,2,4,power_law_1.01,0.32144256591796877
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,2,4,power_law_1.01,0.3231155014038086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,2,4,power_law_1.01,0.3401267242431641
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,2,4,power_law_1.01,0.3596160125732422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,2,4,power_law_1.01,0.3529331207275391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,2,4,power_law_1.01,0.3561177444458008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,2,4,power_law_1.01,0.3906982421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,2,4,power_law_1.01,0.4186393737792969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,2,4,power_law_1.01,0.47083263397216796
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,2,4,balanced,0.9541030120849608
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,2,4,power_law_1.01,0.5161523056030274
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,2,4,power_law_1.01,0.66102783203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,2,4,power_law_1.01,0.035676159858703614
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,2,4,power_law_1.01,0.8065459442138672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,2,4,power_law_1.01,0.050184960365295406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,2,4,power_law_1.01,0.07675392150878906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,2,4,power_law_1.01,1.1419519805908203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,2,4,power_law_1.01,0.08745344161987305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,2,4,power_law_1.01,0.10278783798217775
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,2,4,power_law_1.01,0.12804863929748536
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,2,4,power_law_1.01,0.1371660804748535
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,2,4,power_law_1.01,0.12399999618530273
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,2,4,power_law_1.01,0.13888511657714844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,2,4,power_law_1.01,1.4897715759277343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,2,4,power_law_1.01,0.1418175983428955
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,2,4,power_law_1.01,0.14292096138000487
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,2,4,power_law_1.01,0.14766207695007325
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,2,4,power_law_1.01,0.15370112419128418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,2,4,power_law_1.01,0.15117055892944337
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,2,4,power_law_1.01,0.1562713623046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,2,4,power_law_1.01,0.061489920616149905
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,2,4,power_law_1.01,2.1219879150390626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,2,4,power_law_1.01,0.2084121513366699
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,2,4,power_law_1.01,0.24690303802490235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,2,4,power_law_1.01,0.2814412879943847
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,2,4,power_law_1.01,0.35818241119384764
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,2,4,power_law_1.01,0.4315903854370117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,2,4,power_law_1.01,2.814355163574219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,2,4,power_law_1.01,0.5716966247558594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,2,4,power_law_1.2,0.0478323221206665
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,2,4,power_law_1.01,0.7406566619873047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,2,4,power_law_1.01,0.16482944488525392
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,2,4,power_law_1.2,0.07062016010284425
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,2,4,power_law_1.2,0.09101056098937989
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,2,4,power_law_1.01,0.18939775466918946
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,2,4,power_law_1.2,0.10163711547851562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,2,4,power_law_1.2,0.16047615051269531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,2,4,power_law_1.2,0.17726079940795897
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,2,4,power_law_1.01,1.0761138916015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,2,4,power_law_1.2,0.31183744430541993
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,2,4,power_law_1.2,0.309486083984375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,2,4,power_law_1.2,0.31867008209228515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,2,4,power_law_1.01,1.3900338745117187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,2,4,power_law_1.2,0.3288755035400391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,2,4,power_law_1.2,0.3275187301635742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,2,4,power_law_1.2,0.3319001770019531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,2,4,power_law_1.2,0.3443724822998047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,2,4,power_law_1.2,0.3641548919677734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,2,4,power_law_1.2,0.3572671890258789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,2,4,power_law_1.2,0.40208000183105475
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,2,4,power_law_1.2,0.19921407699584961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,2,4,power_law_1.2,0.20489215850830078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,2,4,power_law_1.2,0.41857662200927737
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,2,4,power_law_1.2,0.49025535583496094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,2,4,power_law_1.2,0.5373056030273438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,2,4,power_law_1.2,0.6649510192871093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,2,4,power_law_1.2,0.035487999916076665
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,2,4,power_law_1.2,0.8458214569091798
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,2,4,power_law_1.2,0.05032832145690917
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,2,4,power_law_1.2,0.060392961502075196
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,2,4,power_law_1.2,1.165279998779297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,2,4,power_law_1.2,0.0719167995452881
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,2,4,power_law_1.2,0.08936448097229004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,2,4,power_law_1.2,0.10409215927124023
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,2,4,power_law_1.2,0.11692288398742676
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,2,4,power_law_1.2,1.5647193908691406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,2,4,power_law_1.2,0.12320639610290526
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,2,4,power_law_1.2,0.14028800010681153
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,2,4,power_law_1.2,0.14110207557678223
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,2,4,power_law_1.2,0.14555904388427734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,2,4,power_law_1.2,0.14724608421325686
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,2,4,power_law_1.2,0.15531135559082032
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,2,4,power_law_1.2,0.1453286361694336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,2,4,power_law_1.2,0.15497599601745607
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,2,4,power_law_1.2,0.16072320938110352
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,2,4,power_law_1.2,0.1704217529296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,2,4,power_law_1.2,0.19757440567016601
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,2,4,power_law_1.2,2.315010528564453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,2,4,power_law_1.2,0.21232511520385744
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,2,4,power_law_1.2,0.2527872085571289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,2,4,power_law_1.2,0.28733951568603516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,2,4,power_law_1.2,0.3827609634399414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,2,4,power_law_1.2,0.44960895538330076
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,4,2,balanced,0.03162623882293701
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,2,4,power_law_1.2,0.5959040069580078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,4,2,balanced,0.07373439788818359
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,2,4,power_law_1.2,2.992464599609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,2,4,power_law_1.2,0.7733081817626953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,4,2,balanced,0.23277183532714846
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,2,4,power_law_1.2,1.141514205932617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,4,2,balanced,0.2414566421508789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,4,2,balanced,0.236232967376709
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,2,4,power_law_1.2,1.4676914978027344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,4,2,balanced,0.23817087173461915
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,4,2,balanced,0.24167423248291015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,4,2,balanced,0.24277503967285155
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,4,2,balanced,0.2544716835021973
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,4,2,balanced,0.2576678466796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,4,2,balanced,0.2595340728759766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,4,2,balanced,0.23900415420532228
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,4,2,balanced,0.26378751754760743
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,4,2,balanced,0.2778892707824707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,4,2,balanced,0.3171251106262207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,4,2,balanced,0.3436812973022461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,4,2,balanced,0.37110782623291017
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,4,2,balanced,0.4173158264160156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,4,2,balanced,0.473834228515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,4,2,balanced,0.6963814544677734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,4,2,balanced,0.7955353546142578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,4,2,balanced,0.02327039957046509
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,4,2,balanced,0.026682879924774174
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,4,2,balanced,0.05392384052276611
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,4,2,balanced,0.04685184001922608
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,4,2,balanced,0.08726271629333496
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,4,2,balanced,0.14696191787719726
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,4,2,balanced,0.14946175575256349
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,4,2,balanced,1.4804019165039062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,4,2,balanced,0.14927871704101564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,4,2,balanced,0.15086976051330567
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,4,2,balanced,0.15150848388671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,4,2,balanced,0.03880064010620117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,4,2,balanced,0.1525273609161377
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,4,2,balanced,0.17571199417114258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,4,2,balanced,0.17722368240356445
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,4,2,balanced,1.134845428466797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,4,2,balanced,0.17900672912597657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,4,2,balanced,0.18185087203979494
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,4,2,balanced,0.19048704147338866
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,4,2,balanced,0.2022643280029297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,4,2,balanced,0.21535999298095704
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,4,2,balanced,0.23839744567871093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,4,2,balanced,0.2657510375976563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,4,2,balanced,0.3161612892150879
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,4,2,balanced,0.36464767456054686
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,4,2,balanced,0.12798463821411132
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,4,2,balanced,0.5470246505737305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,4,2,balanced,0.22821376800537108
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,4,2,balanced,0.18425983428955078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,4,2,balanced,0.6471308898925782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,4,2,power_law_1.01,0.04644480228424072
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,4,2,power_law_1.01,0.08313088417053223
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,4,2,balanced,0.9294719696044922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,4,2,power_law_1.01,0.11134976387023925
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,4,2,balanced,0.02822016000747681
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,4,2,balanced,1.2056883239746095
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,4,2,power_law_1.01,0.13128576278686524
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,4,2,power_law_1.01,0.1890764808654785
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,4,2,power_law_1.01,0.19446399688720703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,4,2,power_law_1.01,0.19684864044189454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,4,2,power_law_1.01,0.19825408935546876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,4,2,power_law_1.01,0.20227071762084964
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,4,2,power_law_1.01,0.2078348731994629
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,4,2,power_law_1.01,0.21557119369506833
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,4,2,power_law_1.01,0.2342361640930176
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,4,2,power_law_1.01,0.23396095275878906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,4,2,power_law_1.01,0.2430873680114746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,4,2,power_law_1.01,0.282744312286377
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,4,2,power_law_1.01,0.313756160736084
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,4,2,power_law_1.01,0.3934105682373047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,4,2,power_law_1.01,0.4779904174804687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,4,2,power_law_1.01,0.5860543823242187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,4,2,power_law_1.01,0.749955825805664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,4,2,power_law_1.01,0.02455296039581299
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,4,2,power_law_1.01,0.0375987195968628
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,4,2,power_law_1.01,1.3197952270507813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,4,2,power_law_1.01,0.061514239311218265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,4,2,power_law_1.01,0.04817535877227783
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,4,2,power_law_1.01,0.059006719589233404
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,4,2,power_law_1.01,0.08007935523986817
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,4,2,power_law_1.01,0.03457535982131958
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,4,2,power_law_1.01,0.09386367797851562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,4,2,power_law_1.01,1.6679129028320312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,4,2,power_law_1.01,0.12430335998535155
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,4,2,power_law_1.01,0.12866175651550293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,4,2,power_law_1.01,0.12955136299133302
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,4,2,power_law_1.01,0.13137663841247557
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,4,2,power_law_1.01,0.1568012809753418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,4,2,power_law_1.01,0.15982080459594727
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,4,2,power_law_1.01,0.16501760482788086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,4,2,power_law_1.01,0.17554304122924805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,4,2,power_law_1.01,0.17265151977539062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,4,2,power_law_1.01,0.1773427200317383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,4,2,power_law_1.01,0.1954572868347168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,4,2,power_law_1.01,0.21870975494384765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,4,2,power_law_1.01,0.12408960342407227
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,4,2,power_law_1.01,0.284719352722168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,4,2,power_law_1.01,0.3574220657348633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,4,2,power_law_1.01,0.4308287811279297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,4,2,power_law_1.01,0.5782617568969727
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,4,2,power_law_1.01,0.7240921783447265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,4,2,power_law_1.01,0.25175167083740235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,4,2,power_law_1.2,0.04568704128265381
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,4,2,power_law_1.01,1.0052787017822264
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,4,2,power_law_1.2,0.05883903980255127
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,4,2,power_law_1.2,0.07876736164093018
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,4,2,power_law_1.2,0.108787202835083
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,4,2,power_law_1.2,0.12361599922180175
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,4,2,power_law_1.01,1.3092889404296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,4,2,power_law_1.2,0.17852031707763671
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,4,2,power_law_1.2,0.19656831741333008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,4,2,power_law_1.2,0.20456064224243162
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,4,2,power_law_1.2,0.21092735290527345
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,4,2,power_law_1.2,0.21983104705810547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,4,2,power_law_1.01,0.2404979133605957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,4,2,power_law_1.2,0.23662464141845704
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,4,2,power_law_1.2,0.24
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,4,2,power_law_1.2,0.2492710494995117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,4,2,power_law_1.2,0.28746112823486325
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,4,2,power_law_1.2,0.3291443252563476
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,4,2,power_law_1.2,0.36361473083496093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,4,2,power_law_1.2,0.4085849761962891
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,4,2,power_law_1.2,0.4863078308105469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,4,2,power_law_1.2,0.6005939102172851
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,4,2,power_law_1.2,0.03449088096618652
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,4,2,power_law_1.2,0.7726541137695312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,4,2,power_law_1.01,0.35929729461669924
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,4,2,power_law_1.2,1.0188915252685546
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,4,2,power_law_1.2,0.024462080001831053
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,4,2,power_law_1.2,0.03807487964630127
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,4,2,power_law_1.2,0.04670976161956787
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,4,2,power_law_1.2,1.3871372985839845
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,4,2,power_law_1.2,0.056989440917968745
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,4,2,power_law_1.2,0.07942143917083741
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,4,2,power_law_1.01,0.956737289428711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,4,2,power_law_1.2,0.08677887916564941
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,4,2,power_law_1.2,0.12359935760498046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,4,2,power_law_1.2,1.7305690002441405
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,4,2,power_law_1.2,0.12729855537414553
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,4,2,power_law_1.2,0.1295257568359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,4,2,power_law_1.2,0.13582207679748534
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,4,2,power_law_1.2,0.1551257610321045
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,4,2,power_law_1.2,0.19784576416015626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,4,2,power_law_1.2,0.19207040786743163
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,4,2,power_law_1.2,0.16607744216918946
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,4,2,power_law_1.2,0.1696767997741699
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,4,2,power_law_1.2,0.1729395294189453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,4,2,power_law_1.2,0.18109312057495117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,4,2,power_law_1.2,0.20345727920532228
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,4,2,power_law_1.2,0.229671688079834
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,4,2,power_law_1.2,0.2574489593505859
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,4,2,power_law_1.2,0.12182656288146973
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,4,2,power_law_1.2,0.2977382469177246
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,4,2,power_law_1.2,0.37300094604492184
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,4,2,power_law_1.2,0.4365478515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,4,2,power_law_1.2,0.230696964263916
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,4,2,power_law_1.2,0.16098304748535158
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,8,1,balanced,0.033068161010742184
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,8,1,balanced,0.04500224113464356
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,8,1,balanced,0.05830592155456543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,8,1,balanced,0.08083200454711914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,8,1,balanced,0.13547327995300293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,8,1,balanced,0.2454649543762207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,8,1,balanced,0.24367807388305662
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,8,1,balanced,0.2474380874633789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,8,1,balanced,0.2485126495361328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,8,1,balanced,0.2508492851257324
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,8,1,balanced,0.2532211112976074
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,4,2,power_law_1.2,1.3502796936035157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,8,1,balanced,0.25665279388427736
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,8,1,balanced,0.25985408782958985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,4,2,power_law_1.2,0.7478246307373047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,8,1,balanced,0.2653683280944824
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,8,1,balanced,0.35255870819091795
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,8,1,balanced,0.35856639862060546
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,8,1,balanced,0.3683065414428711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,4,2,power_law_1.2,1.033697280883789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,8,1,balanced,0.38795455932617184
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,8,1,balanced,0.4004019165039063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,8,1,balanced,0.4306854248046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,4,2,power_law_1.2,0.5902233505249024
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,8,1,balanced,0.4638150405883789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,8,1,balanced,0.6130956649780274
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,8,1,balanced,0.03152575969696045
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,8,1,balanced,0.6782067108154297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,8,1,balanced,0.04924543857574463
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,8,1,balanced,0.07109439849853516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,8,1,balanced,0.9262822723388672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,8,1,balanced,0.11485823631286621
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,8,1,balanced,0.19408063888549804
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,8,1,balanced,0.19998207092285156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,8,1,balanced,0.1989414405822754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,8,1,balanced,1.1959526062011718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,8,1,balanced,0.1986131286621094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,8,1,balanced,0.20159744262695312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,8,1,balanced,0.2030201530456543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,8,1,balanced,0.2088819122314453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,8,1,balanced,0.20612224578857422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,8,1,balanced,1.7511013793945314
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,8,1,balanced,0.2577491188049316
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,8,1,balanced,0.2619987106323242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,8,1,balanced,0.26800384521484377
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,8,1,balanced,0.19608320236206053
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,8,1,balanced,0.28247360229492186
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,8,1,balanced,0.30243200302124024
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,8,1,balanced,0.32192127227783207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,8,1,balanced,0.3634905624389648
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,8,1,balanced,2.2755967712402345
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,8,1,balanced,0.036823680400848394
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,8,1,power_law_1.01,0.029906558990478515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,8,1,balanced,0.25668224334716794
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,8,1,balanced,0.6460934448242187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,8,1,balanced,0.7423833465576172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,8,1,power_law_1.01,0.05801087856292725
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,8,1,power_law_1.01,0.08113984107971192
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,8,1,power_law_1.01,0.11146240234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,8,1,balanced,0.4086265563964844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,8,1,balanced,1.0766899108886718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,8,1,power_law_1.01,0.1435795211791992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,8,1,power_law_1.01,0.20675392150878907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,8,1,power_law_1.01,0.21071680068969725
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,8,1,power_law_1.01,0.2141945648193359
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,8,1,balanced,1.3946745300292969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,8,1,power_law_1.01,0.2156800079345703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,8,1,power_law_1.01,0.04652736186981201
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,8,1,power_law_1.01,0.2230739212036133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,8,1,power_law_1.01,0.23154815673828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,8,1,power_law_1.01,0.23642112731933596
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,8,1,power_law_1.01,0.25960575103759764
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,8,1,power_law_1.01,0.3309734344482422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,8,1,power_law_1.01,0.3334425735473633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,8,1,power_law_1.01,0.34648128509521486
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,8,1,power_law_1.01,0.36680255889892577
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,8,1,power_law_1.01,0.3896710586547852
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,8,1,power_law_1.01,0.45643966674804687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,8,1,power_law_1.01,0.4924403381347656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,8,1,power_law_1.01,0.6018963241577149
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,8,1,power_law_1.01,0.03690752029418946
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,8,1,power_law_1.01,0.7254911804199219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,8,1,power_law_1.01,0.04898752212524414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,8,1,power_law_1.01,0.9913075256347657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,8,1,power_law_1.01,0.07042496204376221
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,8,1,power_law_1.01,0.11896256446838378
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,8,1,power_law_1.01,1.2664588928222655
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,8,1,power_law_1.01,0.15701696395874024
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,8,1,power_law_1.01,0.16289600372314453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,8,1,power_law_1.01,0.16642944335937498
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,8,1,power_law_1.01,0.1699078369140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,8,1,power_law_1.01,0.17414400100708008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,8,1,power_law_1.01,0.17818431854248046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,8,1,power_law_1.01,1.8191098022460939
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,8,1,power_law_1.01,0.18070207595825197
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,8,1,power_law_1.01,0.18471040725708007
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,8,1,power_law_1.01,0.23595775604248045
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,8,1,power_law_1.01,0.23833728790283204
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,8,1,power_law_1.01,2.3742764282226565
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,8,1,power_law_1.01,0.25955072402954105
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,8,1,power_law_1.01,0.27856191635131833
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,8,1,power_law_1.01,0.322380142211914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,8,1,power_law_1.01,0.35714366912841794
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,8,1,power_law_1.01,0.43551486968994146
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,8,1,power_law_1.01,0.031554560661315914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,8,1,power_law_1.01,0.2467673683166504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,8,1,power_law_1.01,0.6823961639404297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,8,1,power_law_1.2,0.028813440799713135
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,8,1,power_law_1.2,0.046776318550109865
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,8,1,power_law_1.2,0.058568959236145024
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,8,1,power_law_1.01,0.863289566040039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,8,1,power_law_1.2,0.0773644781112671
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,8,1,power_law_1.01,0.5176095962524414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,8,1,power_law_1.2,0.10502335548400879
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,8,1,power_law_1.01,1.1906297302246094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,8,1,power_law_1.2,0.205479679107666
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,8,1,power_law_1.2,0.20968191146850587
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,8,1,power_law_1.2,0.2122483253479004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,8,1,power_law_1.2,0.21688064575195312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,8,1,power_law_1.2,0.22106943130493165
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,8,1,power_law_1.2,0.22886720657348633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,8,1,power_law_1.01,0.1002732753753662
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,8,1,power_law_1.01,1.5305715942382814
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,8,1,power_law_1.2,0.2690912055969238
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,8,1,power_law_1.2,0.3320671844482422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,8,1,power_law_1.2,0.3392377471923828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,8,1,power_law_1.2,0.3493747329711914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,8,1,power_law_1.2,0.13921279907226564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,8,1,power_law_1.2,0.37672000885009765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,8,1,power_law_1.2,0.3991923141479492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,8,1,power_law_1.2,0.4764556884765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,8,1,power_law_1.2,0.505928955078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,8,1,power_law_1.2,0.23643968582153319
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,8,1,power_law_1.2,0.6191001510620117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,8,1,power_law_1.2,0.7524877166748046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,8,1,power_law_1.2,0.03156160116195679
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,8,1,power_law_1.2,0.03793407917022705
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,8,1,power_law_1.2,0.04918399810791015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,8,1,power_law_1.2,1.0071501159667968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,8,1,power_law_1.2,0.06837376117706298
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,8,1,power_law_1.2,0.09097791671752929
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,8,1,power_law_1.2,0.11469183921813966
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,8,1,power_law_1.2,0.15346112251281738
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,8,1,power_law_1.2,1.2899180603027343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,8,1,power_law_1.2,0.16226879119873047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,8,1,power_law_1.2,0.1648134422302246
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,8,1,power_law_1.2,0.16758975982666016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,8,1,power_law_1.2,0.17252992630004882
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,8,1,power_law_1.2,0.18705087661743164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,8,1,power_law_1.2,0.23887487411499025
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,8,1,power_law_1.2,1.8309651184082032
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,8,1,power_law_1.2,0.24339391708374025
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,8,1,power_law_1.2,0.24897727966308594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,8,1,power_law_1.2,0.27172031402587893
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,8,1,power_law_1.2,0.2999052810668945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,8,1,power_law_1.2,0.3347340774536133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,8,1,power_law_1.2,0.36842689514160154
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,8,1,power_law_1.2,0.17657983779907227
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,8,1,power_law_1.2,2.3849900817871097
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,8,1,power_law_1.2,0.4523712158203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,1,8,balanced,0.02565887928009033
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,8,1,power_law_1.2,0.5395417785644531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,1,8,balanced,0.042978558540344235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,1,8,balanced,0.04812032222747803
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,1,8,balanced,0.10417407989501952
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,1,8,balanced,0.17832960128784178
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,8,1,power_law_1.2,0.69569091796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,1,8,balanced,0.18106111526489257
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,1,8,balanced,0.18244735717773436
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,1,8,balanced,0.18731136322021485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,8,1,power_law_1.2,0.8768121337890624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,1,8,balanced,0.1912704086303711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,1,8,balanced,0.2025177574157715
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,1,8,balanced,0.20756479263305666
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,1,8,balanced,0.19843072891235353
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,1,8,balanced,0.20432256698608398
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,8,1,power_law_1.2,1.2118463897705078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,1,8,balanced,0.20537599563598632
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,1,8,balanced,0.2137766456604004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,1,8,balanced,0.23608320236206054
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,1,8,balanced,0.2554534339904785
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,8,1,power_law_1.2,1.5409152221679687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,1,8,balanced,0.3102400016784668
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,1,8,balanced,0.3332620620727539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,1,8,balanced,0.18966144561767578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,1,8,balanced,0.4440959930419922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,8,1,power_law_1.2,0.1799443244934082
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,1,8,balanced,0.5054438400268555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,1,8,power_law_1.01,0.061734399795532226
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,1,8,power_law_1.01,0.09090175628662109
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,1,8,power_law_1.01,0.11336064338684082
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,1,8,balanced,0.929516830444336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,1,8,power_law_1.01,0.10144000053405762
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,1,8,power_law_1.01,0.1518950366973877
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,1,8,balanced,1.3613247680664062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,1,8,power_law_1.01,0.16114303588867188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,1,8,power_law_1.01,0.16960384368896483
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,1,8,balanced,0.0700390386581421
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,1,8,power_law_1.01,0.17628927230834962
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,1,8,power_law_1.01,0.1853286361694336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,1,8,balanced,1.7672064208984373
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,1,8,power_law_1.01,0.19434495925903322
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,1,8,balanced,0.7493734741210938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,1,8,power_law_1.01,0.20289279937744142
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,1,8,power_law_1.01,0.21166080474853516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,1,8,power_law_1.01,0.2101273536682129
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,1,8,power_law_1.01,0.22545536041259764
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,1,8,power_law_1.01,0.23603328704833984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,1,8,power_law_1.01,0.24637311935424805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,1,8,power_law_1.01,0.26739967346191407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,1,8,power_law_1.01,0.29467519760131833
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,1,8,power_law_1.01,0.37440769195556645
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,1,8,power_law_1.01,0.42020736694335936
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,1,8,power_law_1.01,0.6109247970581054
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,1,8,power_law_1.01,0.7280563354492188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,1,8,power_law_1.2,0.06188159942626953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,1,8,power_law_1.2,0.07023359775543213
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,1,8,power_law_1.2,0.09266559600830078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,1,8,power_law_1.2,0.10288895606994627
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,1,8,power_law_1.2,0.10173184394836425
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,1,8,power_law_1.01,1.4209689331054687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,1,8,power_law_1.2,0.15552127838134766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,1,8,power_law_1.01,0.09613311767578125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,1,8,power_law_1.2,0.17696128845214845
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,1,8,power_law_1.2,0.18630399703979492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,1,8,power_law_1.2,0.18375295639038086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,1,8,power_law_1.2,0.1964134407043457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,1,8,power_law_1.2,0.2083673667907715
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,1,8,power_law_1.2,0.21148416519165042
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,1,8,power_law_1.2,0.2152704048156738
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,1,8,power_law_1.2,0.23083263397216797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,1,8,power_law_1.01,2.1920191955566404
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,1,8,power_law_1.01,1.1263423919677735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,1,8,power_law_1.2,0.24398975372314452
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,1,8,power_law_1.2,0.25287168502807617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,1,8,power_law_1.2,0.2773030471801758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,1,8,power_law_1.2,0.31083391189575194
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,1,8,power_law_1.2,0.16499584197998046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,1,8,power_law_1.2,0.3915903854370117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,1,8,power_law_1.2,0.5114828872680663
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,2,4,balanced,0.022536959648132324
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,1,8,power_law_1.2,0.6504179382324219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,1,8,power_law_1.01,3.0992141723632813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,2,4,balanced,0.03327744007110596
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,2,4,balanced,0.0443123197555542
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,2,4,balanced,0.06895743846893311
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,2,4,balanced,0.111080961227417
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,2,4,balanced,0.16617984771728517
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,2,4,balanced,0.1978175926208496
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,2,4,balanced,0.1985830307006836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,1,8,power_law_1.2,1.2825395202636718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,2,4,balanced,0.2016896057128906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,2,4,balanced,0.2040985679626465
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,2,4,balanced,0.20903167724609376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,2,4,balanced,0.2155289649963379
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,2,4,balanced,0.20626176834106444
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,2,4,balanced,0.21231359481811524
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,1,8,power_law_1.2,1.6650253295898438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,2,4,balanced,0.21063552856445314
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,2,4,balanced,0.21693183898925783
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,2,4,balanced,0.22470016479492189
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,1,8,power_law_1.2,0.840390396118164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,2,4,balanced,0.2468953514099121
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,2,4,balanced,0.2652300834655762
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,2,4,balanced,0.3037094306945801
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,2,4,balanced,0.3327052688598633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,2,4,balanced,0.44550529479980466
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,2,4,balanced,0.5109056091308594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,2,4,power_law_1.01,0.043843841552734374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,1,8,power_law_1.2,2.649939270019531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,2,4,balanced,0.74980224609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,2,4,power_law_1.01,0.054606080055236816
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,2,4,power_law_1.01,0.06830848217010498
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,2,4,balanced,0.9356441497802734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,2,4,power_law_1.01,0.17087871551513673
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,2,4,power_law_1.01,0.18270208358764647
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,2,4,balanced,1.3572889709472657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,1,8,power_law_1.2,3.5139672851562502
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,2,4,power_law_1.01,0.18962432861328124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,2,4,power_law_1.01,0.19575040817260742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,2,4,power_law_1.01,0.20464639663696288
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,2,4,power_law_1.01,0.21359487533569338
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,2,4,balanced,1.7720600891113283
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,2,4,power_law_1.01,0.2165171241760254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,2,4,power_law_1.01,0.21475584030151368
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,2,4,power_law_1.01,0.22651903152465822
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,2,4,power_law_1.01,0.24298112869262695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,2,4,power_law_1.01,0.26091007232666014
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,2,4,power_law_1.01,0.2682393646240234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,2,4,power_law_1.01,0.31794944763183597
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,2,4,power_law_1.01,0.37993473052978516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,2,4,power_law_1.01,0.08413311958312988
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,2,4,power_law_1.01,0.08889984130859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,2,4,power_law_1.01,0.12480383872985841
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,2,4,power_law_1.01,0.48016384124755856
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,2,4,power_law_1.01,0.2076915168762207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,2,4,power_law_1.2,0.043668479919433595
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,2,4,power_law_1.01,0.6122009658813476
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,2,4,power_law_1.2,0.04974207878112793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,2,4,power_law_1.2,0.06669824123382569
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,2,4,power_law_1.2,0.09220992088317871
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,2,4,power_law_1.01,1.0285670471191406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,2,4,power_law_1.2,0.17252351760864257
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,2,4,power_law_1.01,1.1423910522460938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,2,4,power_law_1.2,0.18869760513305664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,2,4,power_law_1.2,0.19431167602539062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,2,4,power_law_1.2,0.19522176742553712
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,2,4,power_law_1.01,1.6257843017578124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,2,4,power_law_1.2,0.20843263626098635
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,2,4,power_law_1.2,0.21486719131469725
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,2,4,power_law_1.2,0.21171199798583987
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,2,4,power_law_1.2,0.21590656280517578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,2,4,power_law_1.2,0.22266752243041993
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,2,4,power_law_1.2,0.23636991500854493
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,2,4,power_law_1.2,0.2523072052001953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,2,4,power_law_1.01,2.2677516174316406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,2,4,power_law_1.2,0.267646713256836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,2,4,power_law_1.2,0.2800281524658203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,2,4,power_law_1.2,0.07197696208953858
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,2,4,power_law_1.2,0.33648769378662113
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,2,4,power_law_1.2,0.40452735900878906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,2,4,power_law_1.2,0.13684864044189454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,2,4,power_law_1.2,0.5358348846435547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,4,2,balanced,0.020362238883972168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,4,2,balanced,0.03128319978713989
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,4,2,balanced,0.04338304042816162
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,2,4,power_law_1.2,0.6780595397949218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,4,2,balanced,0.06599936008453369
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,4,2,balanced,0.10366592407226563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,4,2,balanced,0.16352767944335939
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,4,2,balanced,0.18747903823852538
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,4,2,balanced,0.18930688858032227
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,4,2,balanced,0.19094911575317383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,2,4,power_law_1.2,1.284005126953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,4,2,balanced,0.19398143768310547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,4,2,balanced,0.19657600402832032
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,4,2,balanced,0.20463232040405271
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,4,2,balanced,0.20286592483520507
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,4,2,balanced,0.20731647491455077
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,4,2,balanced,0.2148147201538086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,4,2,balanced,0.23251583099365236
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,4,2,balanced,0.16773632049560547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,4,2,balanced,0.24647552490234376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,4,2,balanced,0.2981900787353516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,4,2,balanced,0.32885375976562503
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,4,2,balanced,0.4509734344482422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,4,2,balanced,0.5237004852294922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,2,4,power_law_1.2,1.1444338989257812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,4,2,balanced,0.771704330444336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,4,2,power_law_1.01,0.030327041149139405
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,4,2,power_law_1.01,0.04123136043548584
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,4,2,balanced,0.9614169311523437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,4,2,power_law_1.01,0.052567038536071774
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,4,2,power_law_1.01,0.06701824188232422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,4,2,power_law_1.01,0.11546239852905274
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,2,4,power_law_1.2,2.4273829650878906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,4,2,power_law_1.01,0.13937664031982422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,4,2,power_law_1.01,0.1400499153137207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,4,2,balanced,1.3939878845214844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,2,4,power_law_1.2,1.8571353149414063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,4,2,power_law_1.01,0.1695859146118164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,4,2,power_law_1.01,0.18276735305786135
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,4,2,power_law_1.01,0.17900287628173828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,4,2,power_law_1.01,0.18836864471435547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,4,2,balanced,1.8230860900878905
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,4,2,balanced,0.1673356819152832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,4,2,power_law_1.01,0.235468807220459
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,4,2,power_law_1.01,0.0808140754699707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,4,2,power_law_1.01,0.17428224563598632
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,4,2,power_law_1.01,0.19737728118896486
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,4,2,power_law_1.01,0.2087295913696289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,4,2,power_law_1.01,0.46065406799316405
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,4,2,power_law_1.01,0.24814592361450197
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,4,2,power_law_1.01,0.5728204727172852
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,4,2,power_law_1.01,0.29949951171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,4,2,power_law_1.01,0.3647884750366211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,4,2,power_law_1.2,0.030778880119323733
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,4,2,power_law_1.01,0.8761087799072266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,4,2,power_law_1.2,0.03966207981109619
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,4,2,power_law_1.2,0.051938557624816896
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,4,2,power_law_1.01,0.2006924819946289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,4,2,power_law_1.2,0.06269055843353272
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,4,2,power_law_1.2,0.07901951789855957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,4,2,power_law_1.01,1.0555213165283202
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,4,2,power_law_1.2,0.11940863609313965
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,4,2,power_law_1.01,0.26186880111694333
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,4,2,power_law_1.2,0.14172672271728515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,4,2,power_law_1.2,0.14219648361206055
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,4,2,power_law_1.2,0.17389823913574218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,4,2,power_law_1.2,0.17355775833129883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,4,2,power_law_1.2,0.18572160720825195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,4,2,power_law_1.2,0.18844287872314452
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,4,2,power_law_1.01,1.523493194580078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,4,2,power_law_1.2,0.20436223983764648
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,4,2,power_law_1.2,0.2107097625732422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,4,2,power_law_1.2,0.22322816848754884
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,4,2,power_law_1.2,0.25170560836791994
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,4,2,power_law_1.2,0.255993595123291
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,4,2,power_law_1.2,0.26583679199218746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,4,2,power_law_1.2,0.30529920578002934
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,4,2,power_law_1.01,2.0704025268554687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,4,2,power_law_1.2,0.3734707260131836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,4,2,power_law_1.2,0.4786995315551758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,4,2,power_law_1.2,0.5869990539550781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,4,2,power_law_1.2,0.19432832717895507
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,4,2,power_law_1.2,1.0874918365478516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,4,2,power_law_1.2,2.1550630187988284
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,4,2,power_law_1.2,0.9152973175048829
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,8,1,balanced,0.030259840488433838
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,8,1,balanced,0.04711359977722168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,8,1,balanced,0.06855936050415039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,8,1,balanced,0.10373056411743165
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,4,2,power_law_1.2,1.6128230285644531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,8,1,balanced,0.17907583236694335
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,8,1,balanced,0.25987968444824217
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,8,1,balanced,0.26436735153198243
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,8,1,balanced,0.2633158493041992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,8,1,balanced,0.2635487937927246
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,8,1,balanced,0.2648230361938476
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,8,1,balanced,0.2663884735107422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,8,1,balanced,0.3013395118713379
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,8,1,balanced,0.5170572662353516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,8,1,balanced,0.3015692710876465
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,8,1,balanced,0.3087833595275879
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,8,1,balanced,0.30740352630615236
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,8,1,balanced,0.5330752182006836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,8,1,balanced,0.3133139228820801
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,8,1,balanced,0.3191935920715332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,8,1,balanced,0.3339039993286133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,8,1,balanced,0.7363616180419922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,8,1,power_law_1.01,0.029662721157073975
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,8,1,balanced,0.7923321533203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,8,1,balanced,0.34406593322753903
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,8,1,power_law_1.01,0.06863935947418212
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,8,1,balanced,1.2231078338623047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,8,1,power_law_1.01,0.12340864181518554
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,8,1,power_law_1.01,0.1819296073913574
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,8,1,power_law_1.01,0.19271232604980468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,8,1,power_law_1.01,0.21026304244995114
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,8,1,balanced,1.495401611328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,8,1,power_law_1.01,0.22501056671142577
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,8,1,power_law_1.01,0.23689407348632813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,8,1,power_law_1.01,0.25639551162719726
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,8,1,power_law_1.01,0.27167552947998047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,8,1,power_law_1.01,0.28065664291381837
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,8,1,power_law_1.01,0.29171840667724613
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,8,1,balanced,2.1799052429199217
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,8,1,power_law_1.01,0.30059583663940426
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,8,1,power_law_1.01,0.3118252754211426
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,8,1,power_law_1.01,0.35974399566650395
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,8,1,power_law_1.01,0.0490342378616333
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,8,1,power_law_1.01,0.09499584197998047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,8,1,power_law_1.01,0.3846598434448242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,8,1,power_law_1.01,0.4114451217651367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,8,1,balanced,2.854760437011719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,8,1,power_law_1.01,0.4705171203613281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,8,1,power_law_1.01,0.5592499160766602
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,8,1,power_law_1.01,0.7131654357910155
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,8,1,power_law_1.01,0.8857497406005859
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,8,1,power_law_1.2,0.06799168109893798
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,8,1,power_law_1.2,0.08834495544433593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,8,1,power_law_1.01,1.2315647888183594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,8,1,power_law_1.01,1.5693760681152344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,8,1,power_law_1.2,0.21900480270385742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,8,1,power_law_1.2,0.029669120311737063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,8,1,power_law_1.2,0.049452800750732426
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,8,1,power_law_1.2,0.23191488265991209
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,8,1,power_law_1.01,2.2479417419433596
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,8,1,power_law_1.2,0.1969273567199707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,8,1,power_law_1.2,0.24588224411010745
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,8,1,power_law_1.2,0.281790714263916
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,8,1,power_law_1.2,0.28938367843627927
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,8,1,power_law_1.2,0.11708928108215333
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,8,1,power_law_1.2,0.3073747253417969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,8,1,power_law_1.2,0.3204972839355469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,8,1,power_law_1.01,2.926622619628906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,8,1,power_law_1.2,0.18193151473999022
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,8,1,power_law_1.2,0.3469676971435547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,8,1,power_law_1.2,0.3695705413818359
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,8,1,power_law_1.2,0.39182849884033205
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,8,1,power_law_1.2,0.4154764938354492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,8,1,power_law_1.2,0.48182144165039065
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,8,1,power_law_1.2,0.5632108688354492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,8,1,power_law_1.2,0.7249517059326172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,1,8,balanced,0.07534719944000244
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,8,1,power_law_1.2,0.8896511840820314
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,1,8,balanced,0.1331443214416504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,8,1,power_law_1.2,0.2605964851379395
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,1,8,balanced,0.4613312149047852
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,1,8,balanced,0.7567475128173828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,8,1,power_law_1.2,1.237408676147461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,1,8,balanced,1.017976303100586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,1,8,balanced,1.0244198608398438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,1,8,balanced,1.0303513336181642
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,8,1,power_law_1.2,1.5809913635253907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,1,8,balanced,1.5394342041015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,1,8,balanced,1.5444915771484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,1,8,balanced,0.25918720245361326
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,1,8,balanced,1.5608563232421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,1,8,balanced,1.583072052001953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,1,8,balanced,1.6011264038085937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,1,8,balanced,1.6447654724121095
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,1,8,balanced,1.6560140991210937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,1,8,balanced,1.7050994873046874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,8,1,power_law_1.2,2.2664793395996092
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,1,8,balanced,1.764031982421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,1,8,balanced,1.9185920715332032
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,1,8,balanced,2.022818603515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,1,8,balanced,2.252500457763672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,1,8,balanced,2.454895324707031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,1,8,balanced,0.05270400047302246
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,1,8,balanced,0.09611392021179199
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,1,8,balanced,3.550307922363281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,1,8,balanced,0.24383232116699222
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,1,8,balanced,4.118678894042969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,1,8,balanced,0.39524097442626954
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,1,8,balanced,0.5273459243774414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,8,1,power_law_1.2,2.9425445556640626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,1,8,balanced,0.5377484893798827
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,1,8,balanced,0.5445862579345703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,1,8,balanced,0.5817113494873046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,1,8,balanced,0.5849766540527344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,1,8,balanced,0.592883186340332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,1,8,balanced,6.523239746093751
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,1,8,balanced,0.6064627075195312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,1,8,balanced,0.6226227188110351
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,1,8,balanced,0.6564812469482422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,1,8,balanced,0.6439411163330078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,1,8,balanced,0.6589504241943359
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,1,8,balanced,0.701891860961914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,1,8,balanced,0.7626802825927734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,1,8,balanced,0.8264128112792969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,1,8,balanced,8.278975830078124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,1,8,balanced,0.9615436553955078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,1,8,balanced,1.0894681549072265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,1,8,balanced,1.5665370178222655
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,1,8,balanced,0.14017279624938966
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,1,8,balanced,1.8510899353027344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,1,8,power_law_1.01,0.37444862365722653
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,1,8,balanced,2.8439486694335936
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,1,8,power_law_1.01,0.7036569976806641
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,1,8,power_law_1.01,0.6213619232177734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,1,8,balanced,3.160350646972656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,1,8,power_law_1.01,0.7416639709472657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,1,8,power_law_1.01,0.8641446685791015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,1,8,power_law_1.01,0.9459814453125001
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,1,8,balanced,12.890374755859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,1,8,balanced,4.854384765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,1,8,power_law_1.01,1.5085708618164062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,1,8,balanced,6.118367919921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,1,8,power_law_1.01,0.8933644866943359
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,1,8,power_law_1.01,0.8830630493164062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,1,8,balanced,16.22094482421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,1,8,power_law_1.01,1.532725830078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,1,8,power_law_1.01,1.3146701049804688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,1,8,power_law_1.01,1.3216371154785156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,1,8,power_law_1.01,1.4031053161621094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,1,8,power_law_1.01,1.6675596618652342
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,1,8,power_law_1.01,1.9571865844726564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,1,8,power_law_1.01,2.1214707946777347
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,1,8,power_law_1.01,2.5419827270507813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,1,8,power_law_1.01,3.0743487548828123
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,1,8,power_law_1.01,1.5882893371582032
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,1,8,power_law_1.01,1.605445098876953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,1,8,power_law_1.01,3.9376422119140626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,1,8,power_law_1.01,0.19840768814086912
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,1,8,power_law_1.01,1.7545408630371093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,1,8,power_law_1.01,0.3621926498413086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,1,8,power_law_1.01,0.39106048583984376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,1,8,power_law_1.01,0.4574694442749023
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,1,8,power_law_1.01,5.020711669921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,1,8,power_law_1.01,6.7637451171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,1,8,power_law_1.01,8.98830078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,1,8,power_law_1.01,0.3208281707763672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,1,8,power_law_1.01,0.503430404663086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,1,8,power_law_1.01,0.47137920379638676
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,1,8,power_law_1.01,0.4667443084716797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,1,8,power_law_1.01,14.434345703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,1,8,power_law_1.01,17.003225097656248
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,1,8,power_law_1.01,0.4913919830322266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,1,8,power_law_1.01,0.6666470336914062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,1,8,power_law_1.01,0.7483916473388672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,1,8,power_law_1.01,0.4974476623535156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,1,8,power_law_1.01,0.8383526611328126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,1,8,power_law_1.01,0.5425715255737305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,1,8,power_law_1.01,1.014750747680664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,1,8,power_law_1.01,1.190123519897461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,1,8,power_law_1.01,0.5654758453369141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,1,8,power_law_1.01,1.6243455505371096
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,1,8,power_law_1.01,0.6226943969726563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,1,8,power_law_1.01,1.9904780578613281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,1,8,power_law_1.01,0.5800588989257813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,1,8,power_law_1.2,0.3773376083374024
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,1,8,power_law_1.01,2.7900173950195315
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,1,8,power_law_1.2,0.5187583923339844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,1,8,power_law_1.01,0.6043532943725586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,1,8,power_law_1.01,3.5922125244140624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,1,8,power_law_1.2,0.742155532836914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,1,8,power_law_1.2,0.9421491241455078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,1,8,power_law_1.2,0.9719385528564454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,1,8,power_law_1.2,0.8567922973632813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,1,8,power_law_1.01,5.1207958984375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,1,8,power_law_1.01,0.6249843215942382
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,1,8,power_law_1.2,1.3481829833984373
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,1,8,power_law_1.2,1.3672012329101562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,1,8,power_law_1.2,1.4312947082519532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,1,8,power_law_1.2,0.640505599975586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,1,8,power_law_1.2,1.5279399108886718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,1,8,power_law_1.2,1.5671347045898438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,1,8,power_law_1.01,7.026160888671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,1,8,power_law_1.2,1.652007751464844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,1,8,power_law_1.2,1.7046911621093748
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,1,8,power_law_1.2,1.7786688232421874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,1,8,power_law_1.2,1.830303955078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,1,8,power_law_1.2,1.9342771911621093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,1,8,power_law_1.2,0.9004914855957031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,1,8,power_law_1.2,2.251790008544922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,1,8,power_law_1.2,2.5912576293945313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,1,8,power_law_1.2,0.20271488189697268
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,1,8,power_law_1.2,3.2660519409179685
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,1,8,power_law_1.2,0.26956031799316404
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,1,8,power_law_1.2,4.19449462890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,1,8,power_law_1.2,0.38830078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,1,8,power_law_1.2,0.4403839874267578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,1,8,power_law_1.2,0.5020275115966797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,1,8,power_law_1.2,5.52590576171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,1,8,power_law_1.2,0.46894592285156256
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,1,8,power_law_1.2,0.4738291168212891
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,1,8,power_law_1.2,0.504582405090332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,1,8,power_law_1.2,0.5357503890991211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,1,8,power_law_1.2,0.554353904724121
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,1,8,power_law_1.2,0.5719795227050781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,1,8,power_law_1.2,0.5941043090820313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,1,8,power_law_1.2,0.637657585144043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,1,8,power_law_1.2,8.090355224609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,1,8,power_law_1.2,0.6172940826416016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,1,8,power_law_1.2,0.6318707275390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,1,8,power_law_1.2,0.6752588653564453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,1,8,power_law_1.2,0.7631948852539063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,1,8,power_law_1.2,0.8707481384277344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,1,8,power_law_1.2,1.031379165649414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,1,8,power_law_1.2,1.2792499542236329
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,1,8,power_law_1.2,10.031182250976562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,1,8,power_law_1.2,0.3237798309326172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,1,8,power_law_1.2,1.7514866638183595
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,2,4,balanced,0.07601408004760743
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,2,4,balanced,0.11571711540222167
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,1,8,power_law_1.2,2.149996795654297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,2,4,balanced,0.24293376922607424
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,2,4,balanced,0.40178176879882815
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,1,8,power_law_1.2,3.0220440673828124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,2,4,balanced,0.7182195281982422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,2,4,balanced,1.0076172637939453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,2,4,balanced,1.0145254516601563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,2,4,balanced,1.0232077026367188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,1,8,power_law_1.2,4.185390014648438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,2,4,balanced,1.0386163330078124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,2,4,balanced,1.045955810546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,2,4,balanced,1.058551025390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,2,4,balanced,1.5386904907226562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,2,4,balanced,1.5549478149414062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,1,8,power_law_1.2,5.716404418945312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,2,4,balanced,1.5996070861816407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,2,4,balanced,1.6015513610839842
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,1,8,power_law_1.2,16.109715576171872
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,2,4,balanced,1.6506906127929688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,2,4,balanced,1.7144755554199218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,2,4,balanced,1.8621784973144533
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,2,4,balanced,1.9707443237304687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,2,4,balanced,2.195063018798828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,2,4,balanced,2.394220733642578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,2,4,balanced,3.3252801513671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,2,4,balanced,0.053094401359558105
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,1,8,power_law_1.2,21.08409423828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,2,4,balanced,3.7890393066406247
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,2,4,balanced,0.08292480468749999
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,2,4,balanced,0.1338918399810791
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,2,4,balanced,0.21690879821777345
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,1,8,power_law_1.2,7.474329833984375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,2,4,balanced,0.383950080871582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,2,4,balanced,0.527088623046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,2,4,balanced,0.5389132690429688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,2,4,balanced,0.5575244903564454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,2,4,balanced,0.5507072067260742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,2,4,balanced,0.5615808105468749
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,2,4,balanced,0.5717440032958985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,2,4,balanced,0.6289267349243164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,2,4,balanced,0.6393932723999024
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,2,4,balanced,6.014111938476562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,2,4,balanced,0.6803993225097656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,2,4,balanced,0.6619161224365234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,2,4,balanced,0.6715763092041016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,2,4,balanced,0.71162109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,2,4,balanced,0.7667340850830078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,2,4,balanced,0.8264780426025391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,2,4,balanced,7.59704345703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,2,4,balanced,0.9165849304199218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,2,4,balanced,1.0264793395996095
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,2,4,balanced,1.5130636596679687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,2,4,balanced,1.6914151000976563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,2,4,power_law_1.01,0.2350783920288086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,2,4,power_law_1.01,0.37346817016601563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,2,4,power_law_1.01,0.3881651306152344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,2,4,balanced,2.929912414550781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,2,4,power_law_1.01,0.5167564773559571
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,2,4,power_law_1.01,0.6379199981689453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,2,4,balanced,2.596741027832031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,2,4,balanced,11.4557861328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,2,4,power_law_1.01,0.8378112030029297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,2,4,balanced,4.4064001464843745
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,2,4,power_law_1.01,0.8845811462402343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,2,4,power_law_1.01,0.9103218841552735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,2,4,power_law_1.01,0.885191650390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,2,4,power_law_1.01,0.9462553405761719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,2,4,power_law_1.01,1.4207577514648437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,2,4,power_law_1.01,1.4310169982910157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,2,4,power_law_1.01,1.470786590576172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,2,4,balanced,5.59236083984375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,2,4,power_law_1.01,1.5458688354492187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,2,4,power_law_1.01,1.6131135559082033
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,2,4,balanced,14.871782226562498
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,2,4,power_law_1.01,1.702406463623047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,2,4,power_law_1.01,1.8507379150390626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,2,4,power_law_1.01,1.9598835754394532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,2,4,power_law_1.01,0.9389939117431642
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,2,4,power_law_1.01,2.2950810241699218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,2,4,power_law_1.01,2.727829895019531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,2,4,power_law_1.01,0.12348671913146973
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,2,4,power_law_1.01,0.2054591941833496
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,2,4,power_law_1.01,3.408043518066406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,2,4,power_law_1.01,0.27140991210937504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,2,4,power_law_1.01,0.3381747055053711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,2,4,power_law_1.01,0.4922649765014649
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,2,4,power_law_1.01,4.481551208496094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,2,4,power_law_1.01,0.450379524230957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,2,4,power_law_1.01,0.46648830413818365
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,2,4,power_law_1.01,0.5010585784912109
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,2,4,power_law_1.01,5.914580688476563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,2,4,power_law_1.01,0.5772864151000976
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,2,4,power_law_1.01,0.5867622375488282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,2,4,power_law_1.01,0.6430348968505859
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,2,4,power_law_1.01,7.575827026367188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,2,4,power_law_1.01,0.6354393768310547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,2,4,power_law_1.01,0.6510617828369141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,2,4,power_law_1.01,0.4837516784667969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,2,4,power_law_1.01,0.6882701110839844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,2,4,power_law_1.01,0.8678745269775391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,2,4,power_law_1.01,0.2015283203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,2,4,power_law_1.01,0.5217177581787109
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,2,4,power_law_1.01,0.9727833557128907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,2,4,power_law_1.01,1.167837448120117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,2,4,power_law_1.01,1.492236785888672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,2,4,power_law_1.01,1.7840039062500002
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,2,4,power_law_1.01,11.640943603515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,2,4,power_law_1.2,0.2330419158935547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,2,4,power_law_1.2,0.27916160583496097
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,2,4,power_law_1.2,0.38659713745117186
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,2,4,power_law_1.2,0.5017203140258789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,2,4,power_law_1.01,3.2669082641601563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,2,4,power_law_1.01,2.499201354980469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,2,4,power_law_1.2,0.6018803024291992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,2,4,power_law_1.2,0.8431692504882813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,2,4,power_law_1.2,0.8592575836181642
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,2,4,power_law_1.2,0.8678284454345704
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,2,4,power_law_1.2,0.8849830627441406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,2,4,power_law_1.01,0.7484864044189453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,2,4,power_law_1.01,4.5281701660156255
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,2,4,power_law_1.2,0.8977101135253907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,2,4,power_law_1.01,15.52834716796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,2,4,power_law_1.2,0.9404032135009766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,2,4,power_law_1.2,1.4042048645019531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,2,4,power_law_1.2,1.4561970520019532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,2,4,power_law_1.2,1.5772787475585939
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,2,4,power_law_1.2,1.568423614501953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,2,4,power_law_1.2,1.6082534790039062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,2,4,power_law_1.01,6.23807373046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,2,4,power_law_1.2,1.7137510681152341
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,2,4,power_law_1.2,1.8961048889160157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,2,4,power_law_1.2,2.002618865966797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,2,4,power_law_1.2,2.356317443847656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,2,4,power_law_1.2,0.12396160125732422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,2,4,power_law_1.2,2.850330810546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,2,4,power_law_1.2,0.16056959152221678
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,2,4,power_law_1.2,0.21109376907348634
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,2,4,power_law_1.2,3.566824951171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,2,4,power_law_1.2,0.26206464767456056
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,2,4,power_law_1.2,0.31692544937133793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,2,4,power_law_1.2,0.4426956939697265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,2,4,power_law_1.2,0.45467647552490237
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,2,4,power_law_1.2,4.5550079345703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,2,4,power_law_1.2,0.4622182464599609
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,2,4,power_law_1.2,0.5058905410766601
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,2,4,power_law_1.2,0.5314995193481445
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,2,4,power_law_1.2,0.5913216018676758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,2,4,power_law_1.2,0.613359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,2,4,power_law_1.2,6.606231079101562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,2,4,power_law_1.2,0.6535462188720703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,2,4,power_law_1.2,0.6330508804321289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,2,4,power_law_1.2,0.6444326019287109
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,2,4,power_law_1.2,7.801804809570312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,2,4,power_law_1.2,0.6975411224365234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,2,4,power_law_1.2,0.7391193389892579
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,2,4,power_law_1.2,0.8476518249511719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,2,4,power_law_1.2,0.4935436630249024
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,2,4,power_law_1.2,1.002350082397461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,2,4,power_law_1.2,1.2043814086914062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,2,4,power_law_1.2,1.495863037109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,4,2,balanced,0.07414656162261964
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,4,2,balanced,0.11920639991760254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,2,4,power_law_1.2,1.9239859008789062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,4,2,balanced,0.23907968521118167
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,4,2,balanced,0.3996044921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,2,4,power_law_1.2,2.6995162963867188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,4,2,balanced,0.7161510467529297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,2,4,power_law_1.2,3.442291259765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,4,2,balanced,1.019384307861328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,4,2,balanced,1.0393087768554687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,4,2,balanced,1.0304473876953124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,2,4,power_law_1.2,12.4566357421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,4,2,balanced,1.0525580596923827
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,4,2,balanced,1.0712230682373047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,2,4,power_law_1.2,4.716220092773438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,4,2,balanced,1.1117273712158204
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,4,2,balanced,1.13797119140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,4,2,balanced,1.0596556854248047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,4,2,balanced,1.092584991455078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,4,2,balanced,1.588299560546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,4,2,balanced,1.6223718261718751
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,4,2,balanced,1.6788505554199218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,4,2,balanced,1.8062655639648437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,4,2,balanced,1.9085311889648438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,4,2,balanced,2.186282196044922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,2,4,power_law_1.2,16.590627441406248
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,4,2,balanced,2.3495960998535157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,4,2,balanced,0.05241856098175048
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,4,2,balanced,0.07848447799682617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,4,2,balanced,3.221510314941406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,4,2,balanced,0.125099515914917
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,4,2,balanced,0.21951488494873045
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,4,2,balanced,3.6656256103515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,4,2,balanced,0.5504870223999023
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,2,4,power_law_1.2,6.506793212890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,4,2,balanced,0.5648883056640626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,4,2,balanced,0.3873907089233398
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,4,2,balanced,0.5639334487915039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,4,2,balanced,0.5781670379638671
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,4,2,balanced,0.5789555358886719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,4,2,balanced,0.5891609573364258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,4,2,balanced,0.6036723327636719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,4,2,balanced,0.6107865524291992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,4,2,balanced,0.6323161697387695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,4,2,balanced,5.613616943359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,4,2,balanced,0.7131430053710938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,4,2,balanced,0.7403596496582031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,4,2,balanced,0.746956787109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,4,2,balanced,0.8065408325195312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,4,2,balanced,0.8496934509277343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,4,2,balanced,6.986654663085938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,4,2,balanced,0.9426367950439453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,4,2,balanced,1.0390962982177734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,4,2,balanced,1.67233154296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,4,2,balanced,1.862388458251953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,4,2,power_law_1.01,0.11153535842895508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,4,2,power_law_1.01,0.269358081817627
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,4,2,power_law_1.01,0.23916288375854494
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,4,2,balanced,2.641534729003906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,4,2,power_law_1.01,0.41471614837646487
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,4,2,power_law_1.01,0.540584945678711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,4,2,balanced,3.394837646484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,4,2,power_law_1.01,0.8488499450683594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,4,2,balanced,10.90381591796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,4,2,power_law_1.01,0.8957427215576171
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,4,2,power_law_1.01,0.9174169921875001
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,4,2,power_law_1.01,0.9044236755371093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,4,2,balanced,4.571052856445313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,4,2,power_law_1.01,0.8561894226074219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,4,2,power_law_1.01,0.9706380462646484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,4,2,power_law_1.01,1.0192614746093749
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,4,2,power_law_1.01,1.0722713470458984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,4,2,power_law_1.01,1.1647296142578125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,4,2,balanced,5.792157592773437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,4,2,power_law_1.01,1.5075750732421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,4,2,power_law_1.01,1.5329446411132812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,4,2,balanced,13.42291259765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,4,2,power_law_1.01,1.6192358398437499
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,4,2,power_law_1.01,1.7410585021972658
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,4,2,power_law_1.01,1.8613311767578125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,4,2,power_law_1.01,2.1469850158691406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,4,2,power_law_1.01,0.07654272079467773
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,4,2,power_law_1.01,2.508830718994141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,4,2,power_law_1.01,0.12639616012573243
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,4,2,power_law_1.01,3.0814297485351565
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,4,2,power_law_1.01,0.15471232414245606
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,4,2,power_law_1.01,0.22949760437011718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,4,2,power_law_1.01,0.2888281631469726
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,4,2,power_law_1.01,3.966475524902344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,4,2,power_law_1.01,0.44814334869384764
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,4,2,power_law_1.01,0.4554431915283203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,4,2,power_law_1.01,0.4907430267333984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,4,2,power_law_1.01,0.4953408050537109
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,4,2,power_law_1.01,0.5072396850585937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,4,2,power_law_1.01,5.545859985351562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,4,2,power_law_1.01,0.5214835357666016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,4,2,power_law_1.01,0.5423923110961913
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,4,2,power_law_1.01,0.5568268966674805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,4,2,power_law_1.01,0.5734860610961914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,4,2,power_law_1.01,0.6836531066894531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,4,2,power_law_1.01,0.6945407867431641
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,4,2,power_law_1.01,0.7223859405517578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,4,2,power_law_1.01,7.02062744140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,4,2,power_law_1.01,0.7975103759765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,4,2,power_law_1.01,0.8648140716552735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,4,2,power_law_1.01,1.0306009674072265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,4,2,power_law_1.01,1.1927999877929687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,4,2,power_law_1.01,1.5419635009765624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,4,2,power_law_1.01,2.1160435485839844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,4,2,power_law_1.01,2.471338195800781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,4,2,power_law_1.01,10.47667236328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,4,2,power_law_1.2,0.26501888275146485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,4,2,power_law_1.2,0.10871295928955078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,4,2,power_law_1.01,3.2884530639648437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,4,2,power_law_1.2,0.49687934875488277
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,4,2,power_law_1.2,0.17275648117065429
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,4,2,power_law_1.2,0.7775129699707032
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,4,2,power_law_1.2,0.8349388885498048
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,4,2,power_law_1.2,0.8924095916748047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,4,2,power_law_1.01,4.6747198486328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,4,2,power_law_1.2,0.9176691436767579
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,4,2,power_law_1.2,0.39175296783447267
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,4,2,power_law_1.2,0.9216844940185547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,4,2,power_law_1.2,0.9705305480957032
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,4,2,power_law_1.2,1.039705581665039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,4,2,power_law_1.01,13.41829345703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,4,2,power_law_1.2,1.114286117553711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,4,2,power_law_1.2,1.241591033935547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,4,2,power_law_1.01,6.118814697265625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,4,2,power_law_1.2,1.5325849914550782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,4,2,power_law_1.2,1.5866636657714843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,4,2,power_law_1.2,1.6650534057617186
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,4,2,power_law_1.2,1.810058288574219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,4,2,power_law_1.2,1.9290112304687501
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,4,2,power_law_1.2,2.2009869384765626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,4,2,power_law_1.2,0.07655424118041992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,4,2,power_law_1.2,2.547937316894531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,4,2,power_law_1.2,0.11069184303283693
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,4,2,power_law_1.2,0.15976575851440428
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,4,2,power_law_1.2,3.289320983886719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,4,2,power_law_1.2,0.20239488601684572
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,4,2,power_law_1.2,0.2602239990234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,4,2,power_law_1.2,0.4236223983764648
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,4,2,power_law_1.2,4.087215270996094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,4,2,power_law_1.2,0.456360969543457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,4,2,power_law_1.2,0.4646118545532227
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,4,2,power_law_1.2,0.494334716796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,4,2,power_law_1.2,0.5043161773681641
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,4,2,power_law_1.2,0.5298380661010742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,4,2,power_law_1.2,0.5550899124145507
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,4,2,power_law_1.2,0.5702643203735351
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,4,2,power_law_1.2,5.5584564208984375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,4,2,power_law_1.2,0.6064652633666993
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,4,2,power_law_1.2,0.7007974243164062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,4,2,power_law_1.2,0.7207001495361328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,4,2,power_law_1.2,0.7616729736328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,4,2,power_law_1.2,0.8323276519775391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,4,2,power_law_1.2,0.9126707458496094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,4,2,power_law_1.2,1.0749823760986328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,4,2,power_law_1.2,7.0905957031249995
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,4,2,power_law_1.2,1.2335040283203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,4,2,power_law_1.2,1.6121061706542967
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,8,1,balanced,0.09066047668457031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,4,2,power_law_1.2,2.0546188354492188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,8,1,balanced,0.12611071586608885
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,8,1,balanced,0.2758393669128418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,4,2,power_law_1.2,2.6027877807617186
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,8,1,balanced,0.40870655059814454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,4,2,power_law_1.2,3.4615988159179687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,4,2,power_law_1.2,10.107655639648438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,8,1,balanced,1.082449951171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,8,1,balanced,1.0792800140380858
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,8,1,balanced,1.0913196563720704
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,8,1,balanced,0.7476217651367187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,4,2,power_law_1.2,4.662718811035156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,8,1,balanced,1.112108154296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,8,1,balanced,1.1338553619384766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,8,1,balanced,1.2136595153808594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,8,1,balanced,1.1494579315185547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,8,1,balanced,1.162353286743164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,8,1,balanced,1.1840255737304688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,4,2,power_law_1.2,13.7501123046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,8,1,balanced,1.4719807434082033
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,8,1,balanced,1.1884134674072266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,8,1,balanced,1.7812710571289063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,4,2,power_law_1.2,6.3514404296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,8,1,balanced,1.7927641296386718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,8,1,balanced,1.9179820251464843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,8,1,balanced,2.4002706909179685
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,8,1,balanced,2.5983963012695312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,8,1,balanced,0.07077631950378419
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,8,1,balanced,0.09857343673706055
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,8,1,balanced,0.1506547164916992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,8,1,balanced,0.2546784019470215
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,8,1,balanced,0.46139968872070314
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,8,1,balanced,0.658239974975586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,8,1,balanced,3.53806396484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,8,1,balanced,0.6682450866699219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,8,1,balanced,0.6712556457519531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,8,1,balanced,0.6873856353759765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,8,1,balanced,0.6837644958496094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,8,1,balanced,3.982416687011719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,8,1,balanced,0.6903513336181641
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,8,1,balanced,0.703433609008789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,8,1,balanced,0.7083833312988281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,8,1,balanced,0.7310246276855469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,8,1,balanced,0.7167724609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,8,1,balanced,0.7295590209960937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,8,1,balanced,0.7524269104003907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,8,1,balanced,0.9137503814697265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,8,1,balanced,0.9637964630126954
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,8,1,balanced,1.0563654327392578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,8,1,balanced,1.160760955810547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,8,1,balanced,6.1706610107421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,8,1,balanced,1.8668415832519532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,8,1,balanced,2.102557373046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,8,1,balanced,7.663975219726562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,8,1,power_law_1.01,0.12590271949768067
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,8,1,balanced,3.087724914550781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,8,1,balanced,3.3910546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,8,1,power_law_1.01,0.08656255722045898
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,8,1,power_law_1.01,0.3732614517211914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,8,1,balanced,11.579489746093751
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,8,1,power_law_1.01,0.5008857727050782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,8,1,power_law_1.01,0.25575103759765627
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,8,1,power_law_1.01,0.8064972686767579
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,8,1,power_law_1.01,0.8791180419921876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,8,1,power_law_1.01,0.9220909118652344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,8,1,power_law_1.01,0.9482931518554688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,8,1,power_law_1.01,0.9612332916259765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,8,1,power_law_1.01,1.0236646270751952
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,8,1,power_law_1.01,1.0582988739013672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,8,1,balanced,5.465850219726563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,8,1,power_law_1.01,1.1282784271240236
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,8,1,power_law_1.01,1.246534423828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,8,1,power_law_1.01,1.3693881225585938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,8,1,balanced,14.37298095703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,8,1,balanced,6.481740112304688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,8,1,power_law_1.01,1.4935379028320312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,8,1,power_law_1.01,1.7531858825683593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,8,1,power_law_1.01,1.7486553955078126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,8,1,power_law_1.01,1.9108665466308594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,8,1,power_law_1.01,2.1943455505371094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,8,1,power_law_1.01,2.548602294921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,8,1,power_law_1.01,0.0692793607711792
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,8,1,power_law_1.01,0.09676992416381835
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,8,1,power_law_1.01,0.14906368255615235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,8,1,power_law_1.01,3.450897827148437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,8,1,power_law_1.01,0.32181697845458984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,8,1,power_law_1.01,4.154788513183594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,8,1,power_law_1.01,0.5066783905029297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,8,1,power_law_1.01,0.5406854248046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,8,1,power_law_1.01,0.5689503860473633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,8,1,power_law_1.01,0.5856403350830078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,8,1,power_law_1.01,0.5917958450317383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,8,1,power_law_1.01,5.907655639648437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,8,1,power_law_1.01,0.6086630249023438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,8,1,power_law_1.01,0.6309420776367187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,8,1,power_law_1.01,0.6384633636474609
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,8,1,power_law_1.01,0.23356864929199217
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,8,1,power_law_1.01,0.6710822296142578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,8,1,power_law_1.01,0.6747065734863281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,8,1,power_law_1.01,0.7046419525146485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,8,1,power_law_1.01,0.7419737243652345
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,8,1,power_law_1.01,0.9079666900634766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,8,1,power_law_1.01,7.662644653320312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,8,1,power_law_1.01,0.9855340576171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,8,1,power_law_1.01,1.1790054321289063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,8,1,power_law_1.01,1.3457766723632811
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,8,1,power_law_1.01,1.7229344177246095
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,8,1,power_law_1.01,2.1162509155273437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,8,1,power_law_1.2,0.12553024291992188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,8,1,power_law_1.01,2.843855895996094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,8,1,power_law_1.2,0.2555891227722168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,8,1,power_law_1.2,0.08698559761047363
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,8,1,power_law_1.01,10.996091308593751
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,8,1,power_law_1.2,0.33081985473632813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,8,1,power_law_1.01,3.5607086181640626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,8,1,power_law_1.2,0.792258529663086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,8,1,power_law_1.2,0.8787615966796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,8,1,power_law_1.2,0.9144818878173828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,8,1,power_law_1.2,0.9423411560058593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,8,1,power_law_1.2,0.4611219024658203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,8,1,power_law_1.01,5.121375732421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,8,1,power_law_1.2,0.9781708526611329
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,8,1,power_law_1.2,1.0374739074707031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,8,1,power_law_1.2,1.0999488067626952
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,8,1,power_law_1.2,1.1423020935058594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,8,1,power_law_1.2,1.2599922943115236
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,8,1,power_law_1.2,1.4131826782226562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,8,1,power_law_1.01,14.2966455078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,8,1,power_law_1.2,1.5682476806640626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,8,1,power_law_1.2,1.8038073730468749
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,8,1,power_law_1.2,1.794706573486328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,8,1,power_law_1.01,6.754286499023438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,8,1,power_law_1.2,1.9661555480957031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,8,1,power_law_1.2,2.250293121337891
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,8,1,power_law_1.2,2.651177673339844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,8,1,power_law_1.2,0.06951168060302734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,8,1,power_law_1.2,0.09847552299499511
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,8,1,power_law_1.2,0.14911680221557616
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,8,1,power_law_1.2,3.392300109863281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,8,1,power_law_1.2,0.22178239822387696
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,8,1,power_law_1.2,0.2909459114074707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,8,1,power_law_1.2,0.49744510650634766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,8,1,power_law_1.2,4.203633422851563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,8,1,power_law_1.2,0.5405388641357421
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,8,1,power_law_1.2,0.5571955108642579
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,8,1,power_law_1.2,0.5824281692504882
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,8,1,power_law_1.2,0.6027078247070312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,8,1,power_law_1.2,0.6191161727905273
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,8,1,power_law_1.2,0.6391961669921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,8,1,power_law_1.2,0.6550886535644531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,8,1,power_law_1.2,0.6971724700927735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,8,1,power_law_1.2,0.7039494323730469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,8,1,power_law_1.2,5.90991943359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,8,1,power_law_1.2,0.7280723571777343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,8,1,power_law_1.2,0.7821798706054688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,8,1,power_law_1.2,0.9603475189208984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,8,1,power_law_1.2,1.0534918212890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,8,1,power_law_1.2,1.217225570678711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,8,1,power_law_1.2,1.3907795715332032
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,8,1,power_law_1.2,1.7908090209960938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,8,1,power_law_1.2,7.507968750000001
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,2,8,balanced,0.10413951873779297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,2,8,balanced,0.0937996768951416
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,2,8,balanced,0.09341823577880859
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,8,1,power_law_1.2,2.1624986267089845
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,2,8,balanced,0.16388864517211915
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,2,8,balanced,0.16557952880859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,2,8,balanced,0.1651430320739746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,2,8,balanced,0.1657753562927246
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,2,8,balanced,0.16599552154541014
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,2,8,balanced,0.16716672897338866
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,2,8,balanced,0.1674291229248047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,2,8,balanced,0.16810623168945313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,2,8,balanced,0.18416255950927735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,2,8,balanced,0.1855743980407715
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,2,8,balanced,0.18895231246948244
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,2,8,balanced,0.1793075180053711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,2,8,balanced,0.1779199981689453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,8,1,power_law_1.2,2.927161560058594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,2,8,balanced,0.18926719665527342
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,2,8,balanced,0.2218931198120117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,2,8,balanced,0.24768640518188478
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,2,8,balanced,0.32841728210449217
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,2,8,balanced,0.3589683151245117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,2,8,balanced,0.5537459182739257
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,2,8,balanced,0.7070681762695312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,2,8,balanced,1.0361727905273437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,2,8,balanced,1.377313232421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,2,8,balanced,0.057392640113830565
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,8,1,power_law_1.2,3.6472973632812495
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,2,8,balanced,0.057506561279296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,2,8,balanced,0.06074495792388916
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,2,8,balanced,0.07839615821838379
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,2,8,balanced,0.07928832054138184
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,2,8,balanced,0.07899519920349121
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,2,8,balanced,0.0793126392364502
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,2,8,balanced,2.051708221435547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,2,8,balanced,0.0796288013458252
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,2,8,balanced,0.07935743808746339
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,2,8,balanced,0.08210047721862793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,2,8,balanced,0.07933440208435058
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,2,8,balanced,0.08054271697998047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,2,8,balanced,0.08271743774414062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,2,8,balanced,0.08193920135498048
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,2,8,balanced,0.08300800323486328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,2,8,balanced,0.08500096321105957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,2,8,balanced,0.08598143577575683
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,2,8,balanced,0.09413120269775391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,2,8,balanced,0.09948032379150391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,2,8,balanced,0.1283033561706543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,2,8,balanced,0.12898943901062013
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,2,8,balanced,2.718177185058594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,2,8,balanced,0.16542207717895507
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,2,8,balanced,0.20716543197631837
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,2,8,balanced,0.30360960006713866
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,2,8,balanced,0.37170688629150395
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,2,8,power_law_1.01,0.0931276798248291
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,2,8,balanced,0.5659596633911133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,2,8,power_law_1.01,0.1633932876586914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,2,8,power_law_1.01,0.16435199737548828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,2,8,balanced,0.7255219268798828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,2,8,power_law_1.01,0.16590208053588867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,2,8,power_law_1.01,0.1654425621032715
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,2,8,power_law_1.01,0.17075456619262694
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,2,8,power_law_1.01,0.1719756889343262
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,2,8,power_law_1.01,0.1843187141418457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,2,8,power_law_1.01,0.19074687957763672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,2,8,power_law_1.01,0.18340351104736327
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,2,8,power_law_1.01,0.1799590492248535
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,2,8,power_law_1.01,0.1974118423461914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,2,8,power_law_1.01,0.21758975982666015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,2,8,power_law_1.01,0.23262975692749022
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,2,8,power_law_1.01,0.30946176528930663
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,2,8,power_law_1.01,0.2680512046813965
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,2,8,power_law_1.01,0.35608448028564454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,8,1,power_law_1.2,11.20906005859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,2,8,power_law_1.01,0.5036454391479492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,2,8,power_law_1.01,0.6392243194580078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,2,8,power_law_1.01,0.9540326690673828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,2,8,power_law_1.01,1.2297203063964843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,8,1,power_law_1.2,5.165216064453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,2,8,power_law_1.01,2.3056600952148436
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,2,8,power_law_1.01,0.05953407764434815
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,2,8,power_law_1.01,0.08031104087829591
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,2,8,power_law_1.01,1.9638143920898437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,2,8,power_law_1.01,0.07955071926116944
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,2,8,power_law_1.01,0.08359807968139649
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,2,8,power_law_1.01,0.07926144123077392
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,2,8,power_law_1.01,0.08207103729248047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,2,8,power_law_1.01,0.08083456039428712
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,2,8,power_law_1.01,0.08235903739929198
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,2,8,power_law_1.01,0.08274432182312011
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,2,8,power_law_1.01,0.08436351776123047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,2,8,power_law_1.01,0.08568448066711426
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,2,8,power_law_1.01,0.09057920455932618
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,2,8,power_law_1.01,0.09358464241027832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,2,8,power_law_1.01,0.09835519790649414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,2,8,power_law_1.01,0.10665472030639649
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,2,8,power_law_1.01,0.12612863540649416
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,2,8,power_law_1.01,0.1334489631652832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,2,8,power_law_1.01,0.15485823631286622
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,2,8,power_law_1.01,0.24603391647338868
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,2,8,power_law_1.01,0.19468032836914062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,2,8,power_law_1.01,0.3377011108398437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,2,8,power_law_1.01,4.063517456054687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,2,8,power_law_1.01,0.5260172653198242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,2,8,power_law_1.01,4.973738098144532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,2,8,power_law_1.01,0.6690393829345703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,2,8,power_law_1.01,1.3669541931152343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,2,8,power_law_1.2,0.09895551681518555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,2,8,power_law_1.2,0.16651264190673828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,2,8,power_law_1.2,0.16621696472167968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,2,8,power_law_1.2,0.1662182426452637
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,2,8,power_law_1.2,0.1648627281188965
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,2,8,power_law_1.01,1.0744409942626953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,2,8,power_law_1.01,8.440121459960938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,2,8,power_law_1.2,0.16703104019165038
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,2,8,power_law_1.2,0.19007999420166016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,2,8,power_law_1.2,0.18780160903930665
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,2,8,power_law_1.2,0.18516992568969726
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,2,8,power_law_1.2,0.19152639389038087
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,2,8,power_law_1.2,0.18610048294067383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,2,8,power_law_1.2,0.20136831283569334
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,2,8,power_law_1.2,0.2139993667602539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,2,8,power_law_1.2,0.2429465675354004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,8,1,power_law_1.2,6.749642333984374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,8,1,power_law_1.2,14.44884033203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,2,8,power_law_1.2,0.26344959259033207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,2,8,power_law_1.2,0.3189990425109863
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,2,8,power_law_1.2,0.3570470428466797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,2,8,power_law_1.2,0.5331110382080079
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,2,8,power_law_1.01,2.175089874267578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,2,8,power_law_1.2,0.6069644927978516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,2,8,power_law_1.2,0.9699199676513672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,2,8,power_law_1.2,1.3767372131347657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,2,8,power_law_1.2,2.0185279846191406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,2,8,power_law_1.2,0.05825024127960206
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,2,8,power_law_1.2,2.363659515380859
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,2,8,power_law_1.2,0.07838848114013672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,2,8,power_law_1.2,0.08086655616760255
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,2,8,power_law_1.2,0.0780736017227173
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,2,8,power_law_1.2,0.07945983886718751
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,2,8,power_law_1.2,0.07959167957305908
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,2,8,power_law_1.2,0.08117759704589844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,2,8,power_law_1.2,0.08142463684082031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,2,8,power_law_1.2,0.08289919853210449
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,2,8,power_law_1.2,0.08385536193847656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,2,8,power_law_1.2,0.08595199584960937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,2,8,power_law_1.2,0.09061887741088867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,2,8,power_law_1.2,0.09350527763366699
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,2,8,power_law_1.2,0.0987673568725586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,2,8,power_law_1.2,0.10577407836914063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,2,8,power_law_1.2,0.12457471847534181
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,2,8,power_law_1.2,0.1322713565826416
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,2,8,power_law_1.2,0.15421183586120607
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,2,8,power_law_1.2,3.979514770507812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,2,8,power_law_1.2,0.2505023956298828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,2,8,power_law_1.2,0.19880447387695313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,2,8,power_law_1.2,0.36948352813720703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,2,8,power_law_1.01,9.994086303710938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,2,8,power_law_1.2,0.5433612823486328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,2,8,power_law_1.2,4.969377136230468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,2,8,power_law_1.01,2.8272625732421877
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,4,4,balanced,0.0663974380493164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,4,4,balanced,0.06387328147888184
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,4,4,balanced,0.07548160076141357
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,2,8,power_law_1.2,0.7065408325195313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,4,4,balanced,0.1244966411590576
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,4,4,balanced,0.12527359962463377
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,4,4,balanced,0.1252403163909912
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,4,4,balanced,0.12718591690063477
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,4,4,balanced,0.12706048011779786
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,4,4,balanced,0.12901247978210448
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,4,4,balanced,0.1283302402496338
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,4,4,balanced,0.13011967658996582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,4,4,balanced,0.13330304145812988
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,4,4,balanced,0.13483903884887696
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,4,4,balanced,0.13527039527893064
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,4,4,balanced,0.13677568435668946
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,4,4,balanced,0.1382592010498047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,4,4,balanced,0.1444223976135254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,4,4,balanced,0.20061439514160156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,4,4,balanced,0.21235071182250978
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,4,4,balanced,0.3234201431274414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,4,4,balanced,0.376360969543457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,4,4,balanced,0.5546803283691406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,4,4,balanced,0.7155583953857422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,2,8,power_law_1.2,1.0886937713623046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,2,8,power_law_1.2,1.2792396545410156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,4,4,balanced,1.0500454711914062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,4,4,balanced,0.03741312026977539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,4,4,balanced,0.037230079174041746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,4,4,balanced,0.04695040225982666
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,4,4,balanced,1.383252410888672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,4,4,balanced,0.05981567859649658
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,4,4,balanced,0.06278912067413331
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,4,4,balanced,0.06047359943389893
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,4,4,balanced,0.06250239849090576
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,4,4,balanced,0.061377282142639156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,4,4,balanced,0.06218751907348633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,4,4,balanced,0.06256383895874024
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,4,4,balanced,0.06364416122436524
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,4,4,balanced,2.0713267517089844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,4,4,balanced,0.0637721586227417
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,4,4,balanced,0.06514560222625733
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,4,4,balanced,0.067292160987854
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,4,4,balanced,0.06802432060241699
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,4,4,balanced,0.07033215999603272
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,4,4,balanced,0.07540095806121826
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,4,4,balanced,0.08539775848388673
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,4,4,balanced,0.08095999717712402
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,4,4,balanced,0.11067520141601564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,4,4,balanced,0.1269324779510498
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,4,4,balanced,0.1783807945251465
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,4,4,balanced,0.22495616912841795
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,4,4,balanced,2.752613220214844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,4,4,balanced,0.3292736053466797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,4,4,balanced,0.4148134231567383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,4,4,power_law_1.01,0.06195839881896973
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,4,4,power_law_1.01,0.07495808124542236
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,4,4,balanced,0.6188544082641602
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,4,4,power_law_1.01,0.11761792182922363
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,4,4,power_law_1.01,0.12590847969055177
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,4,4,power_law_1.01,0.12780287742614746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,4,4,power_law_1.01,0.12562687873840334
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,4,4,balanced,0.8600614166259766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,4,4,power_law_1.01,0.1283993625640869
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,4,4,power_law_1.01,0.12836864471435547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,4,4,power_law_1.01,0.12984064102172851
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,4,4,power_law_1.01,0.1436774444580078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,4,4,power_law_1.01,0.13828864097595214
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,4,4,power_law_1.01,0.13801728248596193
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,4,4,power_law_1.01,0.13738240242004393
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,4,4,power_law_1.01,0.1572544002532959
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,4,4,power_law_1.01,0.1766054344177246
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,4,4,power_law_1.01,0.21455488204956055
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,4,4,power_law_1.01,0.24555648803710936
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,4,4,power_law_1.01,0.41002113342285157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,4,4,power_law_1.01,0.33978111267089844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,4,4,power_law_1.01,0.6442572784423828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,4,4,power_law_1.01,0.8837721252441406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,4,4,power_law_1.01,1.0963571166992188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,4,4,power_law_1.01,1.4488088989257812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,2,8,power_law_1.2,8.755349731445312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,4,4,power_law_1.01,0.03831295967102051
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,4,4,power_law_1.01,0.046949119567871095
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,4,4,power_law_1.01,0.06196224212646484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,4,4,power_law_1.01,2.2527258300781248
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,4,4,power_law_1.01,0.06032639980316162
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,4,4,power_law_1.01,0.06151040077209473
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,4,4,power_law_1.01,0.06129151821136475
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,4,4,power_law_1.01,0.06344575881958008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,4,4,power_law_1.01,2.9233624267578127
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,4,4,power_law_1.01,0.06325503826141357
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,4,4,power_law_1.01,0.06444032192230224
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,4,4,power_law_1.01,0.07008512020111084
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,4,4,power_law_1.01,0.06693759918212891
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,4,4,power_law_1.01,0.06932991981506348
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,4,4,power_law_1.01,0.07352575778961182
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,4,4,power_law_1.01,0.07517312049865724
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,4,4,power_law_1.01,0.07630080223083496
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,4,4,power_law_1.01,0.08393856048583984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,4,4,power_law_1.01,0.10369536399841309
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,2,8,power_law_1.2,12.6221435546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,4,4,power_law_1.01,0.11028736114501954
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,4,4,power_law_1.01,5.285892944335937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,4,4,power_law_1.01,0.14458368301391603
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,4,4,power_law_1.01,0.19941375732421876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,4,4,power_law_1.01,0.22489599227905271
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,4,4,power_law_1.01,0.3337548828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,4,4,power_law_1.01,4.090097961425782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,4,4,power_law_1.2,0.06441472053527833
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,4,4,power_law_1.2,0.06873216152191161
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,4,4,power_law_1.2,0.11910016059875488
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,4,4,power_law_1.2,0.12444543838500977
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,4,4,power_law_1.01,0.6609881591796876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,4,4,power_law_1.2,0.1255679988861084
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,4,4,power_law_1.01,0.4736691284179687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,4,4,power_law_1.2,0.1307532787322998
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,4,4,power_law_1.2,0.12777088165283204
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,4,4,power_law_1.2,0.12898431777954103
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,4,4,power_law_1.2,0.13483391761779787
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,4,4,power_law_1.2,0.13709312438964844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,4,4,power_law_1.2,0.13541503906250002
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,4,4,power_law_1.2,0.13795583724975585
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,4,4,power_law_1.2,0.1382143974304199
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,4,4,power_law_1.2,0.16080383300781248
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,2,8,power_law_1.2,2.237455291748047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,4,4,power_law_1.2,0.20957824707031253
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,4,4,power_law_1.2,0.26449663162231446
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,2,8,power_law_1.2,3.0326437377929687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,4,4,power_law_1.2,0.33842815399169923
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,4,4,power_law_1.2,0.18708351135253906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,4,4,power_law_1.2,0.4137612915039062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,4,4,power_law_1.01,0.8572748565673829
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,4,4,power_law_1.2,0.6751705932617187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,4,4,power_law_1.01,1.124796142578125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,4,4,power_law_1.2,0.7447411346435546
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,4,4,power_law_1.2,0.0381606388092041
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,4,4,power_law_1.2,1.395135955810547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,4,4,power_law_1.2,0.06038271903991699
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,4,4,power_law_1.2,0.06069632053375244
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,4,4,power_law_1.2,1.0833996582031251
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,4,4,power_law_1.2,0.060062718391418454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,4,4,power_law_1.2,0.06103551864624024
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,4,4,power_law_1.2,0.0625497579574585
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,4,4,power_law_1.2,0.06353792190551757
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,4,4,power_law_1.2,0.0640499210357666
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,4,4,power_law_1.2,0.06507775783538819
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,4,4,power_law_1.2,0.06624639987945556
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,4,4,power_law_1.2,0.06895487785339356
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,4,4,power_law_1.2,0.07014400005340576
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,4,4,power_law_1.2,0.08001664161682129
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,4,4,power_law_1.2,0.0848102378845215
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,4,4,power_law_1.2,0.07468800067901611
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,4,4,power_law_1.2,0.1035852813720703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,4,4,power_law_1.2,0.10989312171936035
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,4,4,power_law_1.2,0.04489471912384033
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,4,4,power_law_1.2,0.18417407989501952
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,4,4,power_law_1.2,2.963225708007813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,4,4,power_law_1.2,0.14989824295043946
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,4,4,power_law_1.2,5.783270263671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,4,4,power_law_1.2,0.23473535537719728
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,4,4,power_law_1.2,0.31901695251464846
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,4,4,power_law_1.01,1.6285824584960937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,8,2,balanced,0.04573952198028565
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,8,2,balanced,0.05444736003875732
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,8,2,balanced,0.0675212812423706
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,8,2,balanced,0.10535039901733398
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,8,2,balanced,0.10609279632568361
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,8,2,balanced,0.10632320404052735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,8,2,balanced,0.10700160026550294
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,8,2,balanced,0.10829440116882325
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,8,2,balanced,0.10900863647460937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,8,2,balanced,0.10987263679504396
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,8,2,balanced,0.11147007942199708
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,8,2,balanced,0.1127359962463379
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,4,4,power_law_1.2,0.4224652862548828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,8,2,balanced,0.11555583953857422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,8,2,balanced,0.11910016059875488
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,8,2,balanced,0.12598015785217284
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,8,2,balanced,0.12829695701599123
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,8,2,balanced,0.12368639945983886
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,8,2,balanced,0.19051136016845702
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,8,2,balanced,0.22002176284790037
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,4,4,power_law_1.2,2.524682312011719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,8,2,balanced,0.3282521438598633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,8,2,balanced,0.3962406539916992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,8,2,balanced,0.5652748870849609
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,8,2,balanced,0.7339788818359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,8,2,balanced,1.063176956176758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,4,4,power_law_1.2,0.6869388580322265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,8,2,balanced,0.029760000705718992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,8,2,balanced,0.03396608114242554
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,8,2,balanced,0.04590335845947265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,4,4,power_law_1.2,1.8047731018066409
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,8,2,balanced,0.051983361244201665
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,8,2,balanced,0.053495039939880376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,8,2,balanced,0.05372799873352051
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,8,2,balanced,0.05490687847137451
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,8,2,balanced,2.0966720581054688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,8,2,balanced,0.05787519931793213
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,8,2,balanced,1.411024627685547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,8,2,balanced,0.056779518127441406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,8,2,balanced,0.05740928173065185
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,8,2,balanced,0.05978879928588867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,8,2,balanced,0.061175041198730476
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,8,2,balanced,0.06567679882049561
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,8,2,balanced,0.055895037651062005
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,8,2,balanced,0.06777088165283203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,8,2,balanced,0.0717964792251587
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,8,2,balanced,0.06575359821319579
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,8,2,balanced,0.08057087898254395
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,8,2,balanced,2.7872869873046873
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,8,2,balanced,0.08914048194885253
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,8,2,balanced,0.1259763240814209
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,8,2,balanced,0.14807168006896973
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,8,2,balanced,0.21635967254638672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,8,2,balanced,0.27354368209838864
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,8,2,balanced,0.39817985534667966
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,8,2,balanced,0.5059660720825195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,8,2,power_law_1.01,0.04865920066833496
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,8,2,power_law_1.01,0.055950078964233395
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,8,2,power_law_1.01,0.054611201286315914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,8,2,balanced,0.7656269073486328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,8,2,power_law_1.01,0.10354175567626953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,8,2,power_law_1.01,0.1042137622833252
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,8,2,power_law_1.01,0.10558848381042481
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,8,2,power_law_1.01,0.10758015632629395
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,8,2,balanced,0.9970047760009766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,8,2,power_law_1.01,0.10824064254760742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,8,2,power_law_1.01,0.10920063972473144
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,8,2,power_law_1.01,0.11107328414916992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,8,2,power_law_1.01,0.11468544006347656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,8,2,power_law_1.01,0.11839872360229493
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,8,2,power_law_1.01,0.12668288230895997
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,8,2,power_law_1.01,0.13600000381469726
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,8,2,power_law_1.01,0.14852992057800293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,8,2,power_law_1.01,0.18695295333862305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,8,2,power_law_1.01,0.29473152160644533
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,8,2,power_law_1.01,0.2612518310546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,8,2,power_law_1.01,0.11477375984191895
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,4,4,power_law_1.2,0.8758515167236329
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,8,2,power_law_1.01,0.41020542144775385
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,8,2,power_law_1.01,1.0086067199707032
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,8,2,power_law_1.01,0.5641907119750977
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,8,2,power_law_1.01,0.03149183988571167
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,8,2,power_law_1.01,0.037806079387664795
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,8,2,power_law_1.01,0.039804160594940186
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,8,2,power_law_1.01,0.051892480850219726
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,8,2,power_law_1.01,0.052261118888854985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,8,2,power_law_1.01,0.7860352325439453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,8,2,power_law_1.01,0.052872958183288576
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,8,2,power_law_1.01,0.05464704036712646
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,8,2,power_law_1.01,0.0545740795135498
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,8,2,power_law_1.01,0.055731201171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,8,2,power_law_1.01,0.0559987211227417
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,8,2,power_law_1.01,0.05753471851348877
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,8,2,power_law_1.01,0.05986815929412842
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,8,2,power_law_1.01,0.061806077957153316
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,8,2,power_law_1.01,0.064683518409729
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,8,2,power_law_1.01,0.06859519958496094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,8,2,power_law_1.01,0.07263999938964844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,8,2,power_law_1.01,0.07746304035186767
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,8,2,power_law_1.01,0.09951744079589844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,8,2,power_law_1.01,0.1503436756134033
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,8,2,power_law_1.01,0.1147532844543457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,8,2,power_law_1.01,0.18855167388916017
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,8,2,power_law_1.01,1.3927142333984375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,8,2,power_law_1.01,0.28814975738525395
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,4,4,power_law_1.2,4.145867614746093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,8,2,power_law_1.01,4.21135498046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,4,4,power_law_1.2,1.2420800018310547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,8,2,power_law_1.2,0.051765761375427245
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,8,2,power_law_1.2,0.05478400230407715
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,8,2,power_law_1.2,0.05404799938201904
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,8,2,power_law_1.2,0.10383359909057617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,8,2,power_law_1.2,0.10431103706359864
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,8,2,power_law_1.2,0.10535424232482911
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,8,2,power_law_1.2,0.10744959831237792
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,8,2,power_law_1.2,0.10835968017578126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,8,2,power_law_1.01,0.3370547103881836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,8,2,power_law_1.2,0.10932607650756836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,8,2,power_law_1.2,0.11098624229431153
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,8,2,power_law_1.01,2.211436767578125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,8,2,power_law_1.2,0.11356672286987304
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,8,2,power_law_1.2,0.11534208297729491
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,8,2,power_law_1.2,0.12034048080444334
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,8,2,power_law_1.2,0.12955904006958008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,8,2,power_law_1.2,0.1417676830291748
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,8,2,power_law_1.2,0.1419980812072754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,8,2,power_law_1.2,0.1899033546447754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,8,2,power_law_1.2,0.254520320892334
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,8,2,power_law_1.01,0.4716582489013672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,8,2,power_law_1.01,2.590773620605469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,8,2,power_law_1.2,0.28738304138183596
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,8,2,power_law_1.2,0.4134400177001953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,8,2,power_law_1.01,1.329390106201172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,8,2,power_law_1.2,0.5140582275390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,8,2,power_law_1.2,0.9812659454345702
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,8,2,power_law_1.2,0.9099993896484374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,8,2,power_law_1.01,0.70866943359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,8,2,power_law_1.2,0.03224191904067993
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,8,2,power_law_1.2,0.038858239650726316
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,8,2,power_law_1.2,0.04049536228179932
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,8,2,power_law_1.2,0.05444352149963379
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,8,2,power_law_1.2,0.0526643180847168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,8,2,power_law_1.2,0.054133758544921876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,8,2,power_law_1.2,0.05412864208221435
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,8,2,power_law_1.2,0.05506432056427002
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,8,2,power_law_1.2,0.05542655944824219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,8,2,power_law_1.2,0.05611392021179199
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,8,2,power_law_1.2,0.058327040672302245
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,8,2,power_law_1.2,0.05945087909698486
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,8,2,power_law_1.2,0.061921281814575194
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,8,2,power_law_1.01,0.8590617370605468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,8,2,power_law_1.2,0.06463232040405273
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,8,2,power_law_1.2,0.06844543933868408
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,8,2,power_law_1.2,0.07139328002929687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,8,2,power_law_1.2,0.07698431968688965
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,8,2,power_law_1.2,0.10252672195434571
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,8,2,power_law_1.2,0.1482431983947754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,8,2,power_law_1.2,0.11527935981750488
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,8,2,power_law_1.2,0.18842111587524416
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,8,2,power_law_1.2,0.29451904296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,8,2,power_law_1.2,0.4864972686767578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,8,2,power_law_1.2,2.7346905517578124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,16,1,balanced,0.04675839900970459
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,16,1,balanced,0.055441279411315915
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,16,1,balanced,0.0692844820022583
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,16,1,balanced,0.0695904016494751
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,8,2,power_law_1.2,1.6403149414062501
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,16,1,balanced,0.098853120803833
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,16,1,balanced,0.10052160263061523
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,16,1,balanced,0.10284992218017579
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,16,1,balanced,0.10373439788818359
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,16,1,balanced,0.10420928001403809
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,8,2,power_law_1.2,0.32075775146484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,16,1,balanced,0.10601216316223146
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,16,1,balanced,0.10730815887451173
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,8,2,power_law_1.2,1.5368333435058594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,16,1,balanced,0.11470911979675294
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,16,1,balanced,0.12606911659240722
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,16,1,balanced,0.1266150379180908
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,16,1,balanced,0.12890239715576174
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,16,1,balanced,0.16617408752441407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,16,1,balanced,0.2002182388305664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,16,1,balanced,0.30459968566894535
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,16,1,balanced,0.11010111808776854
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,16,1,balanced,0.3509600067138672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,16,1,balanced,0.534535026550293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,16,1,balanced,0.7147103881835937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,16,1,balanced,0.1020902442932129
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,16,1,balanced,1.0329964447021485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,16,1,balanced,0.030707199573516846
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,16,1,balanced,0.03781183958053589
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,16,1,balanced,0.048533759117126464
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,16,1,balanced,0.04897471904754639
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,16,1,balanced,1.3580883789062501
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,16,1,balanced,0.05530752182006836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,16,1,balanced,0.057290239334106444
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,16,1,balanced,0.057100801467895514
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,16,1,balanced,0.05834623813629151
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,16,1,balanced,0.05884672164916992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,16,1,balanced,0.06014272212982178
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,16,1,balanced,0.0616703987121582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,16,1,balanced,2.013982086181641
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,16,1,balanced,0.06274176120758057
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,16,1,balanced,0.06591231822967529
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,16,1,balanced,0.06785215854644774
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,16,1,balanced,0.06986944198608398
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,16,1,balanced,0.06802688121795655
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,16,1,balanced,0.059514241218566896
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,16,1,balanced,0.08609279632568359
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,8,2,power_law_1.2,0.8928217315673829
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,16,1,balanced,0.12619071960449219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,16,1,balanced,2.6871917724609373
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,16,1,balanced,0.14785152435302734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,16,1,balanced,0.21615360260009764
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,16,1,balanced,0.2744095993041992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,16,1,balanced,0.40564926147460934
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,16,1,balanced,0.505852165222168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,16,1,power_law_1.01,0.047147521972656245
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,16,1,power_law_1.01,0.05379519939422608
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,16,1,power_law_1.01,0.058634881973266606
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,16,1,balanced,0.7520934295654297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,16,1,power_law_1.01,0.09906880378723144
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,16,1,power_law_1.01,0.10066240310668946
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,16,1,balanced,0.9701625823974609
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,8,2,power_law_1.2,1.2511641693115234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,16,1,power_law_1.01,0.10182592391967774
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,16,1,power_law_1.01,0.10283519744873046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,16,1,power_law_1.01,0.10397312164306641
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,16,1,power_law_1.01,0.10700415611267089
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,16,1,balanced,0.08031295776367188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,16,1,power_law_1.01,0.11008064270019531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,16,1,power_law_1.01,0.11642687797546387
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,16,1,power_law_1.01,0.1207430362701416
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,16,1,power_law_1.01,0.1236633586883545
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,16,1,power_law_1.01,0.1317683219909668
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,16,1,power_law_1.01,0.14045632362365723
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,16,1,power_law_1.01,0.1843507194519043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,16,1,power_law_1.01,0.10888447761535644
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,16,1,power_law_1.01,0.22347583770751953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,8,2,power_law_1.2,0.6911974334716797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,16,1,power_law_1.01,0.32386753082275394
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,16,1,power_law_1.01,0.0682470417022705
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,16,1,power_law_1.01,0.3994534301757812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,16,1,power_law_1.01,0.03239808082580566
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,16,1,power_law_1.01,0.038759040832519534
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,16,1,power_law_1.01,0.5604915237426757
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,16,1,power_law_1.01,0.04158143997192383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,16,1,power_law_1.01,0.04874879837036133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,16,1,power_law_1.01,0.05538368225097656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,16,1,power_law_1.01,0.05666944026947022
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,16,1,power_law_1.01,0.05732607841491699
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,16,1,power_law_1.01,0.05758272171020508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,16,1,power_law_1.01,0.05842304229736328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,16,1,power_law_1.01,0.05952127933502197
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,16,1,power_law_1.01,0.06010752201080323
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,16,1,power_law_1.01,0.061338238716125495
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,16,1,power_law_1.01,0.727465591430664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,16,1,power_law_1.01,0.06507840156555175
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,16,1,power_law_1.01,0.07118527889251709
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,16,1,power_law_1.01,0.07307199954986572
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,16,1,power_law_1.01,0.07515071868896485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,16,1,power_law_1.01,0.07382336139678955
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,16,1,power_law_1.01,0.08671680450439453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,8,2,power_law_1.2,4.283674926757813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,16,1,power_law_1.01,0.11421567916870116
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,16,1,power_law_1.01,0.13363903999328614
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,16,1,power_law_1.01,1.051815643310547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,16,1,power_law_1.01,0.1603494453430176
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,16,1,power_law_1.01,0.21948671340942383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,16,1,power_law_1.01,0.28238719940185547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,16,1,power_law_1.2,0.046913280487060546
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,16,1,power_law_1.2,0.053186559677124025
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,16,1,power_law_1.01,1.380723876953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,16,1,power_law_1.2,0.0561081600189209
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,16,1,power_law_1.2,0.09735360145568847
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,16,1,power_law_1.2,0.10000127792358397
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,16,1,power_law_1.2,0.10120960235595704
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,16,1,power_law_1.2,0.10202495574951172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,16,1,power_law_1.2,0.10325247764587402
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,16,1,power_law_1.2,0.10448896408081054
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,16,1,power_law_1.2,0.10814144134521483
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,16,1,power_law_1.01,0.40724864959716794
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,16,1,power_law_1.2,0.11051456451416015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,16,1,power_law_1.2,0.11549247741699217
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,16,1,power_law_1.2,0.11934080123901367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,16,1,power_law_1.2,0.12429120063781737
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,16,1,power_law_1.2,0.06846335887908936
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,16,1,power_law_1.2,0.13368639945983887
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,16,1,power_law_1.2,0.139749755859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,16,1,power_law_1.2,0.17953535079956057
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,16,1,power_law_1.2,0.22750528335571288
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,16,1,power_law_1.01,0.5143231964111328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,16,1,power_law_1.2,0.32109119415283205
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,16,1,power_law_1.01,2.0435910034179687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,16,1,power_law_1.2,0.39942718505859376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,16,1,power_law_1.2,0.5569599914550781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,16,1,power_law_1.01,0.7585286712646484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,16,1,power_law_1.2,0.033656959533691404
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,16,1,power_law_1.2,0.03832767963409424
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,16,1,power_law_1.2,0.03984639883041382
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,16,1,power_law_1.2,0.04872447967529297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,16,1,power_law_1.2,0.055346560478210446
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,16,1,power_law_1.2,0.056608638763427734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,16,1,power_law_1.2,0.05739327907562256
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,16,1,power_law_1.2,0.7198271942138672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,16,1,power_law_1.2,0.05806528091430664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,16,1,power_law_1.2,0.0590496015548706
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,16,1,power_law_1.2,0.05971968173980713
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,16,1,power_law_1.01,2.703460998535156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,16,1,power_law_1.2,0.06039423942565918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,16,1,power_law_1.2,0.06203392028808594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,16,1,power_law_1.2,0.06715519905090332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,16,1,power_law_1.2,0.0710316801071167
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,16,1,power_law_1.2,0.07405119895935058
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,16,1,power_law_1.2,0.07511616230010987
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,16,1,power_law_1.2,0.0733676815032959
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,16,1,power_law_1.2,0.08590271949768066
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,16,1,power_law_1.2,0.1146547222137451
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,16,1,power_law_1.2,0.1330355167388916
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,16,1,power_law_1.01,1.0165907287597657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,16,1,power_law_1.2,0.16062335968017577
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,16,1,power_law_1.2,1.049036178588867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,16,1,power_law_1.2,0.21792959213256835
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,16,1,power_law_1.2,0.2828972816467285
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,16,1,power_law_1.2,1.3795091247558593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,2,8,balanced,0.12442879676818848
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,2,8,balanced,0.12613887786865235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,2,8,balanced,0.12426495552062988
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,2,8,balanced,0.22171392440795898
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,2,8,balanced,0.21804927825927733
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,2,8,balanced,0.22091007232666016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,2,8,balanced,0.21975807189941404
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,2,8,balanced,0.2211903953552246
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,2,8,balanced,0.2228659248352051
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,2,8,balanced,0.22484352111816408
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,2,8,balanced,0.22257280349731445
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,2,8,balanced,0.26220800399780275
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,2,8,balanced,0.2653913688659668
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,2,8,balanced,0.27740543365478515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,2,8,balanced,0.24603519439697266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,2,8,balanced,0.2490291213989258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,2,8,balanced,0.27109888076782224
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,2,8,balanced,0.3329561614990234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,2,8,balanced,0.34623870849609373
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,2,8,balanced,0.5310361480712891
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,2,8,balanced,0.6320422363281251
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,16,1,power_law_1.2,0.4069542312622071
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,2,8,balanced,0.9159372711181641
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,2,8,balanced,1.1932147216796873
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,2,8,balanced,1.7359666442871096
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,2,8,balanced,2.2891917419433594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,2,8,balanced,0.0741593599319458
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,2,8,balanced,0.07298687934875489
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,2,8,balanced,0.07246719837188721
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,2,8,balanced,0.09507328033447265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,2,8,balanced,0.09460351943969728
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,2,8,balanced,0.09547007560729981
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,2,8,balanced,0.09576704025268554
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,2,8,balanced,3.4144857788085936
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,2,8,balanced,0.09585920333862305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,2,8,balanced,0.10079104423522948
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,2,8,balanced,0.09671808242797851
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,2,8,balanced,0.10051584243774414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,2,8,balanced,0.0981107234954834
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,2,8,balanced,0.09932671546936035
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,2,8,balanced,0.0990937614440918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,2,8,balanced,0.10095232009887696
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,2,8,balanced,0.10180480003356933
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,2,8,balanced,0.10505215644836426
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,2,8,balanced,0.11836031913757324
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,2,8,balanced,0.12650367736816406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,2,8,balanced,0.1649126434326172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,2,8,balanced,4.5334603881835935
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,2,8,balanced,0.17035520553588868
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,2,8,balanced,0.26401920318603517
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,2,8,balanced,0.32117504119873047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,2,8,balanced,0.4548096084594727
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,2,8,balanced,0.5635289764404297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,2,8,power_law_1.01,0.13083264350891113
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,2,8,balanced,0.8636441802978515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,2,8,power_law_1.01,0.22057472229003908
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,2,8,power_law_1.01,0.2310041618347168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,2,8,balanced,1.2032012939453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,16,1,power_law_1.2,2.0399372863769534
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,2,8,power_law_1.01,0.22717952728271484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,2,8,power_law_1.01,0.2293094444274902
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,2,8,power_law_1.01,0.23433216094970702
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,2,8,power_law_1.01,0.2588108825683594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,2,8,power_law_1.01,0.26013568878173826
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,2,8,power_law_1.01,0.26797056198120117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,2,8,power_law_1.01,0.2796428871154785
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,2,8,power_law_1.01,0.2561536026000977
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,2,8,power_law_1.01,0.2736345672607422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,2,8,power_law_1.01,0.29222143173217774
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,2,8,power_law_1.01,0.34202880859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,2,8,power_law_1.01,0.491668472290039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,2,8,power_law_1.01,0.4349529647827149
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,2,8,power_law_1.01,0.6140505599975585
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,16,1,power_law_1.2,0.5196812820434571
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,2,8,power_law_1.01,0.8569420623779298
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,2,8,power_law_1.01,1.4689996337890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,2,8,power_law_1.01,1.136696319580078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,2,8,power_law_1.01,2.1841714477539065
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,2,8,power_law_1.01,3.258530578613281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,2,8,power_law_1.01,4.416643981933594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,2,8,power_law_1.01,0.07372928142547608
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,2,8,power_law_1.01,0.09593600273132324
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,2,8,power_law_1.01,0.09638272285461426
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,2,8,power_law_1.01,0.09498751640319825
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,2,8,power_law_1.01,0.09563263893127441
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,2,8,power_law_1.01,0.09653632164001465
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,2,8,power_law_1.01,0.0970956802368164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,2,8,power_law_1.01,0.09791104316711426
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,2,8,power_law_1.01,0.09980928421020507
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,2,8,power_law_1.01,0.10004096031188965
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,2,8,power_law_1.01,0.10288255691528321
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,2,8,power_law_1.01,0.10985088348388672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,2,8,power_law_1.01,0.11034111976623535
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,2,8,power_law_1.01,0.12252032279968261
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,2,8,power_law_1.01,0.14926207542419434
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,16,1,power_law_1.2,0.7593856048583985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,2,8,power_law_1.01,0.1600127983093262
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,2,8,power_law_1.01,0.1758527946472168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,2,8,power_law_1.01,0.22719871520996096
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,2,8,power_law_1.01,7.441973876953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,2,8,power_law_1.01,0.2817036819458008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,2,8,power_law_1.01,0.3737625503540039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,2,8,power_law_1.01,0.8005696105957032
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,2,8,power_law_1.01,6.611253662109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,16,1,power_law_1.2,2.7027883911132813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,2,8,power_law_1.01,0.5419750213623047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,2,8,power_law_1.01,1.053469467163086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,2,8,power_law_1.2,0.12960639953613282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,2,8,power_law_1.2,0.21817983627319334
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,2,8,power_law_1.2,0.22299264907836913
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,2,8,power_law_1.01,1.8949363708496094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,2,8,power_law_1.2,0.23996032714843749
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,2,8,power_law_1.2,0.2254719924926758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,2,8,power_law_1.01,1.7490098571777346
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,2,8,power_law_1.2,0.22471168518066403
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,2,8,power_law_1.2,0.2701696014404297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,2,8,power_law_1.2,0.26020864486694334
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,2,8,power_law_1.2,0.2601702308654785
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,2,8,power_law_1.2,0.2730291175842285
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,2,8,power_law_1.2,0.2607833671569824
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,2,8,power_law_1.2,0.2789913558959961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,2,8,power_law_1.2,0.3084326362609863
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,2,8,power_law_1.2,0.32489471435546874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,2,8,power_law_1.2,0.48744960784912106
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,2,8,power_law_1.2,0.4367142486572265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,2,8,power_law_1.2,0.636819190979004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,2,8,power_law_1.01,3.5877453613281247
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,2,8,power_law_1.2,0.8631206512451172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,2,8,power_law_1.01,4.8904421997070315
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,2,8,power_law_1.2,1.123895034790039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,2,8,power_law_1.2,1.5879129028320311
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,2,8,power_law_1.2,2.017699890136719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,16,1,power_law_1.2,1.0070150756835938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,2,8,power_law_1.2,3.4211160278320314
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,2,8,power_law_1.2,0.07274879932403563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,2,8,power_law_1.2,0.09599743843078613
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,2,8,power_law_1.2,0.09495295524597168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,2,8,power_law_1.2,0.09487360000610352
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,2,8,power_law_1.2,0.09557503700256348
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,2,8,power_law_1.2,0.09661696434020997
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,2,8,power_law_1.2,0.09772159576416015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,2,8,power_law_1.2,0.09910783767700196
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,2,8,power_law_1.2,0.09992575645446777
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,2,8,power_law_1.01,15.2192626953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,2,8,power_law_1.2,0.1008576011657715
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,2,8,power_law_1.2,0.10448767662048339
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,2,8,power_law_1.2,0.11010047912597656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,2,8,power_law_1.2,0.11094400405883789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,2,8,power_law_1.2,0.11887231826782227
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,2,8,power_law_1.2,0.14915072441101074
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,2,8,power_law_1.2,4.2836865234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,2,8,power_law_1.2,0.1676313591003418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,2,8,power_law_1.2,0.25171072006225587
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,2,8,power_law_1.2,0.1539840030670166
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,2,8,power_law_1.2,0.29948799133300785
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,2,8,power_law_1.2,0.4124480056762695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,2,8,power_law_1.2,0.5067494583129883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,2,8,power_law_1.2,0.7213337707519532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,2,8,power_law_1.2,6.457088012695313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,2,8,power_law_1.2,1.0898931121826172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,2,8,power_law_1.2,1.6366297912597656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,4,4,balanced,0.087575044631958
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,4,4,balanced,0.08809727668762207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,4,4,balanced,0.10854016304016115
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,4,4,balanced,0.17341567993164062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,4,4,balanced,0.17203840255737304
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,2,8,power_law_1.2,9.22224365234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,4,4,balanced,0.1731430435180664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,4,4,balanced,0.1786854362487793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,4,4,balanced,0.1743948745727539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,4,4,balanced,0.17685119628906248
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,4,4,balanced,0.17457408905029298
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,4,4,balanced,0.17883520126342772
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,4,4,balanced,0.1840883255004883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,4,4,balanced,0.18721792221069336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,4,4,balanced,0.18618240356445312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,2,8,power_law_1.2,2.0890931701660156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,4,4,balanced,0.18802431106567383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,4,4,balanced,0.18905088424682617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,4,4,balanced,0.20653312683105468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,4,4,balanced,0.3324684906005859
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,4,4,balanced,0.3531315231323242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,4,4,balanced,0.5132582473754883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,4,4,balanced,0.6287667083740234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,4,4,balanced,0.9336013031005859
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,4,4,balanced,1.2013555145263672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,2,8,power_law_1.01,19.684169921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,4,4,balanced,1.7758451843261718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,4,4,balanced,0.05021567821502686
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,4,4,balanced,0.04952960014343262
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,4,4,balanced,0.06324736118316651
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,4,4,balanced,0.07498752117156983
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,4,4,balanced,2.3444505310058594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,4,4,balanced,0.0749951982498169
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,4,4,balanced,0.07742720127105714
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,4,4,balanced,0.0761574411392212
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,4,4,balanced,0.0766323184967041
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,4,4,balanced,0.07674880027770996
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,4,4,balanced,0.0778598403930664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,4,4,balanced,0.07790207862854004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,4,4,balanced,0.0792089605331421
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,4,4,balanced,0.07936895847320556
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,4,4,balanced,0.0828825569152832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,4,4,balanced,0.08377087593078614
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,4,4,balanced,0.08742527961730957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,4,4,balanced,0.09219584465026856
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,4,4,balanced,0.10974080085754394
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,4,4,balanced,3.490126037597656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,4,4,balanced,0.10897791862487793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,4,4,balanced,0.16563327789306642
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,4,4,balanced,0.1954751968383789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,4,4,balanced,0.2645414352416992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,4,4,balanced,0.3276403045654297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,4,4,balanced,0.4946368026733398
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,4,4,balanced,0.6392639923095703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,4,4,balanced,4.679386901855469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,4,4,balanced,0.9965119934082031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,4,4,power_law_1.01,0.0874892807006836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,4,4,power_law_1.01,0.09809920310974121
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,4,4,power_law_1.01,0.17233407974243165
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,4,4,power_law_1.01,0.173306884765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,4,4,balanced,1.3220774841308593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,4,4,power_law_1.01,0.17709312438964844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,4,4,power_law_1.01,0.17497215270996094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,4,4,power_law_1.01,0.18276607513427734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,4,4,power_law_1.01,0.17990655899047853
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,4,4,power_law_1.01,0.19283071517944336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,4,4,power_law_1.01,0.18914176940917968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,4,4,power_law_1.01,0.1917375946044922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,4,4,power_law_1.01,0.2535296058654785
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,4,4,power_law_1.01,0.1788761520385742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,4,4,power_law_1.01,0.2975705528259277
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,4,4,power_law_1.01,0.35750526428222656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,4,4,power_law_1.01,0.4245439910888672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,4,4,power_law_1.01,0.19908992767333983
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,4,4,power_law_1.01,0.5380531311035156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,4,4,power_law_1.01,0.7678809356689453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,4,4,power_law_1.01,0.954260482788086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,4,4,power_law_1.01,1.9730918884277344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,2,8,power_law_1.2,4.503301086425781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,4,4,power_law_1.01,1.2776473236083983
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,2,8,power_law_1.2,15.2142431640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,4,4,power_law_1.01,2.678048095703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,4,4,power_law_1.01,0.04960768222808838
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,4,4,power_law_1.01,0.0576038408279419
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,4,4,power_law_1.01,0.07449215888977051
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,4,4,power_law_1.01,0.07792768001556397
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,4,4,power_law_1.01,0.07581183910369874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,4,4,power_law_1.01,0.07731328010559083
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,4,4,power_law_1.01,0.07706751823425292
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,4,4,power_law_1.01,0.07886464118957519
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,4,4,power_law_1.01,0.0785100793838501
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,4,4,power_law_1.01,0.07929471969604493
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,4,4,power_law_1.01,0.08126208305358887
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,4,4,power_law_1.01,0.08410880088806152
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,4,4,power_law_1.01,0.0855078411102295
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,4,4,power_law_1.01,0.09289471626281738
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,4,4,power_law_1.01,0.10627072334289552
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,4,4,power_law_1.01,0.11332096099853514
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,4,4,power_law_1.01,0.13980928421020508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,4,4,power_law_1.01,0.18117631912231444
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,2,8,power_law_1.2,3.4505355834960936
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,4,4,power_law_1.01,0.24726783752441409
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,4,4,power_law_1.01,0.20676864624023436
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,4,4,power_law_1.01,0.3772518539428711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,4,4,power_law_1.01,0.4756582260131836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,4,4,power_law_1.01,3.8179647827148435
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,4,4,power_law_1.01,0.6792908477783203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,4,4,power_law_1.01,5.54294677734375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,4,4,power_law_1.2,0.0830668830871582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,4,4,power_law_1.2,0.09243264198303222
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,4,4,power_law_1.01,1.0041292572021485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,4,4,power_law_1.2,0.1642572784423828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,4,4,power_law_1.2,0.17576704025268555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,4,4,power_law_1.2,0.17300352096557617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,4,4,power_law_1.2,0.17610111236572265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,4,4,power_law_1.2,0.17543935775756836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,4,4,power_law_1.2,0.17878784179687499
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,4,4,power_law_1.2,0.18201728820800783
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,4,4,power_law_1.2,0.18825344085693357
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,4,4,power_law_1.2,0.18909568786621095
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,4,4,power_law_1.2,0.18840063095092774
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,4,4,power_law_1.2,0.18810367584228516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,4,4,power_law_1.2,0.21647743225097654
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,4,4,power_law_1.2,0.3120230484008789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,4,4,power_law_1.2,0.3341068649291992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,4,4,power_law_1.01,8.112601318359376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,4,4,power_law_1.2,0.4364825439453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,4,4,power_law_1.2,0.5388748931884766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,4,4,power_law_1.01,2.786114501953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,4,4,power_law_1.2,0.9650828552246093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,4,4,power_law_1.2,0.7617984008789063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,4,4,power_law_1.2,1.2980058288574219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,4,4,power_law_1.01,1.2949632263183593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,4,4,power_law_1.2,1.5350579833984375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,4,4,power_law_1.2,2.698525390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,4,4,power_law_1.2,0.04936704158782959
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,4,4,power_law_1.2,0.0556774377822876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,4,4,power_law_1.2,0.07343616008758544
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,4,4,power_law_1.2,0.07494143962860107
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,4,4,power_law_1.2,0.07427711963653565
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,4,4,power_law_1.2,0.07591551780700684
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,4,4,power_law_1.01,2.2906790161132813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,4,4,power_law_1.2,0.076593918800354
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,4,4,power_law_1.2,0.08310400009155273
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,4,4,power_law_1.2,0.07805568218231201
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,4,4,power_law_1.2,0.08387200355529785
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,4,4,power_law_1.2,0.08089088439941407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,4,4,power_law_1.2,0.08529791831970215
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,4,4,power_law_1.2,0.08864255905151366
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,4,4,power_law_1.2,0.08522368431091308
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,4,4,power_law_1.2,0.10798208236694336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,4,4,power_law_1.2,0.11185791969299316
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,4,4,power_law_1.2,0.1342745590209961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,2,8,power_law_1.2,19.580191650390624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,4,4,power_law_1.2,0.16787967681884766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,4,4,power_law_1.2,0.19842815399169922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,4,4,power_law_1.2,0.3401587295532226
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,4,4,power_law_1.2,0.2650982475280762
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,4,4,power_law_1.2,3.6384664916992184
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,4,4,power_law_1.2,0.421233901977539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,4,4,power_law_1.01,11.874622802734375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,4,4,power_law_1.2,5.050983581542969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,8,2,balanced,0.06331776142120361
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,8,2,balanced,0.07346816062927246
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,8,2,balanced,0.09623807907104492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,8,2,balanced,0.14901887893676757
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,8,2,balanced,0.14884480476379394
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,8,2,balanced,0.1504857635498047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,4,4,power_law_1.2,0.6188927841186523
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,8,2,balanced,0.15437567710876465
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,8,2,balanced,0.1555238437652588
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,8,2,balanced,0.15273856163024901
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,8,2,balanced,0.15316479682922363
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,8,2,balanced,0.15755392074584962
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,8,2,balanced,0.1605196762084961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,8,2,balanced,0.16542207717895507
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,8,2,balanced,0.18365568161010742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,8,2,balanced,0.15444864273071288
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,8,2,balanced,0.19007232666015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,8,2,balanced,0.18965248107910154
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,8,2,balanced,0.30385536193847656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,8,2,balanced,0.3584345626831055
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,8,2,balanced,0.5268236923217773
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,8,2,balanced,0.6407398223876953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,8,2,balanced,0.9655808258056642
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,8,2,balanced,1.2162432098388671
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,8,2,balanced,1.809283905029297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,8,2,balanced,2.3911167907714845
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,4,4,power_law_1.2,0.9691442871093751
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,8,2,balanced,0.03820544004440308
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,8,2,balanced,0.04790016174316406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,8,2,balanced,0.06319744110107421
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,8,2,balanced,0.0679859209060669
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,8,2,balanced,0.06885888099670409
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,8,2,balanced,0.06927743911743164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,8,2,balanced,0.07017856121063233
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,8,2,balanced,0.0708351993560791
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,8,2,balanced,3.5843136596679686
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,8,2,balanced,0.0711411190032959
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,8,2,balanced,0.07417471885681153
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,8,2,balanced,0.07423871994018555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,8,2,balanced,0.07790592193603516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,8,2,balanced,0.07963136196136475
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,8,2,balanced,0.08561535835266113
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,8,2,balanced,0.08895999908447265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,4,4,power_law_1.2,1.431953887939453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,8,2,balanced,0.09331199645996094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,8,2,balanced,0.08932479858398437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,8,2,balanced,0.13008383750915525
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,8,2,balanced,0.12465791702270508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,8,2,balanced,0.17809535980224608
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,8,2,balanced,0.22142080307006834
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,8,2,balanced,0.30842880249023436
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,8,2,balanced,0.38793216705322264
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,8,2,balanced,0.5834099197387695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,8,2,balanced,4.77955322265625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,8,2,balanced,0.7652352142333985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,8,2,power_law_1.01,0.06725503921508788
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,8,2,power_law_1.01,0.07573887825012207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,8,2,power_law_1.01,0.08732928276062012
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,8,2,balanced,1.149054718017578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,8,2,power_law_1.01,0.14836480140686034
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,8,2,power_law_1.01,0.1490291213989258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,8,2,power_law_1.01,0.15246848106384275
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,8,2,power_law_1.01,0.153569278717041
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,8,2,power_law_1.01,0.1556172752380371
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,8,2,power_law_1.01,0.15487487792968752
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,8,2,power_law_1.01,0.1545689582824707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,8,2,power_law_1.01,0.15964415550231933
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,8,2,balanced,1.5605068969726563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,8,2,power_law_1.01,0.16880640029907226
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,8,2,power_law_1.01,0.17340288162231446
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,8,2,power_law_1.01,0.19321983337402343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,8,2,power_law_1.01,0.2111871910095215
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,8,2,power_law_1.01,0.3002777671813965
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,8,2,power_law_1.01,0.39613822937011717
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,4,4,power_law_1.2,9.262389526367187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,8,2,power_law_1.01,0.22271743774414063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,8,2,power_law_1.01,0.6832883453369141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,8,2,power_law_1.01,0.4837209701538086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,8,2,power_law_1.01,0.8336013031005859
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,8,2,power_law_1.01,1.1662067413330077
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,4,4,power_law_1.2,8.754365234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,8,2,power_law_1.01,0.043770880699157716
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,8,2,power_law_1.01,0.04841216087341309
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,8,2,power_law_1.01,0.05106815814971923
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,8,2,power_law_1.01,0.06771071910858154
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,8,2,power_law_1.01,0.06819839954376221
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,8,2,power_law_1.01,0.06882559776306152
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,8,2,power_law_1.01,1.549063720703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,8,2,power_law_1.01,0.07053184032440185
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,8,2,power_law_1.01,0.07101183891296386
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,8,2,power_law_1.01,0.07180287837982177
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,8,2,power_law_1.01,0.07228288173675537
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,8,2,power_law_1.01,0.0747699213027954
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,8,2,power_law_1.01,0.07756159782409669
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,8,2,power_law_1.01,0.08021504402160644
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,8,2,power_law_1.01,0.08359295845031738
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,8,2,power_law_1.01,0.0962662410736084
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,8,2,power_law_1.01,0.10084223747253418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,8,2,power_law_1.01,0.10499199867248535
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,8,2,power_law_1.01,0.1340556812286377
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,8,2,power_law_1.01,0.22177408218383787
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,4,4,power_law_1.2,2.6745919799804687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,8,2,power_law_1.01,0.15523584365844728
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,8,2,power_law_1.01,2.4848818969726563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,8,2,power_law_1.01,0.35001598358154296
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,8,2,power_law_1.01,0.2679078483581543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,8,2,power_law_1.01,3.1322445678710937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,4,4,power_law_1.2,2.3823143005371095
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,8,2,power_law_1.2,0.070830078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,8,2,power_law_1.2,0.08027263641357421
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,8,2,power_law_1.2,0.08504192352294922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,8,2,power_law_1.2,0.14914048194885254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,8,2,power_law_1.2,0.14910719871520997
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,8,2,power_law_1.2,0.14933888435363768
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,8,2,power_law_1.2,0.15338879585266113
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,8,2,power_law_1.2,0.15484288215637207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,8,2,power_law_1.2,0.15498368263244627
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,8,2,power_law_1.2,0.15505151748657225
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,8,2,power_law_1.2,0.15997440338134766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,8,2,power_law_1.2,0.16191743850708007
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,8,2,power_law_1.2,0.17235328674316405
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,8,2,power_law_1.2,0.18970752716064454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,8,2,power_law_1.01,0.5156044769287109
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,8,2,power_law_1.2,0.21752063751220704
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,8,2,power_law_1.2,0.20591104507446287
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,8,2,power_law_1.2,0.3004185676574707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,8,2,power_law_1.2,0.38534782409667967
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,8,2,power_law_1.01,4.6702706909179685
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,8,2,power_law_1.01,0.7475839996337891
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,8,2,power_law_1.2,0.6492876434326172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,8,2,power_law_1.2,0.47187969207763675
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,8,2,power_law_1.01,5.870496215820312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,8,2,power_law_1.2,0.8280985260009766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,8,2,power_law_1.01,0.8902105712890626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,8,2,power_law_1.2,1.3826547241210938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,8,2,power_law_1.2,0.0422976016998291
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,8,2,power_law_1.2,0.045863680839538574
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,8,2,power_law_1.2,0.05149951934814453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,8,2,power_law_1.2,0.06735104084014892
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,8,2,power_law_1.2,0.06724607944488525
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,8,2,power_law_1.2,0.06891136169433594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,8,2,power_law_1.2,0.06975615978240966
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,8,2,power_law_1.2,0.0706329584121704
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,8,2,power_law_1.2,0.07177216053009033
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,8,2,power_law_1.2,1.55169921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,8,2,power_law_1.01,1.9473983764648437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,8,2,power_law_1.2,0.07229568004608154
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,8,2,power_law_1.2,0.07478271961212159
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,8,2,power_law_1.01,1.3616383361816407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,8,2,power_law_1.2,0.09030271530151367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,8,2,power_law_1.2,0.0783846378326416
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,8,2,power_law_1.2,0.10502143859863282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,8,2,power_law_1.2,0.08767871856689453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,8,2,power_law_1.2,0.10122240066528321
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,8,2,power_law_1.2,0.09687423706054688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,8,2,power_law_1.2,0.1329817581176758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,8,2,power_law_1.2,0.15382911682128905
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,8,2,power_law_1.2,0.21595392227172852
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,8,2,power_law_1.2,0.2645158386230469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,8,2,power_law_1.2,0.40501377105712894
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,8,2,power_law_1.2,0.45101184844970704
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,8,2,power_law_1.2,0.7055538940429688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,16,1,balanced,0.06333055973052978
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,16,1,balanced,0.07446784019470215
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,8,2,power_law_1.2,2.3858828735351563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,16,1,balanced,0.10149248123168945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,16,1,balanced,0.10312383651733399
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,8,2,power_law_1.2,2.900155029296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,16,1,balanced,0.14493568420410158
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,16,1,balanced,0.1478726387023926
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,16,1,balanced,0.15014975547790527
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,16,1,balanced,0.15018176078796386
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,16,1,balanced,0.15172927856445312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,16,1,balanced,0.1546681594848633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,16,1,balanced,0.1558137607574463
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,16,1,balanced,0.16271488189697264
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,16,1,balanced,0.162346248626709
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,16,1,balanced,0.16824256896972656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,16,1,balanced,0.1841548728942871
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,16,1,balanced,0.18949056625366212
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,16,1,balanced,0.195152645111084
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,16,1,balanced,0.2874566459655762
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,16,1,balanced,0.3149344062805176
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,16,1,balanced,0.4633152008056641
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,16,1,balanced,0.5998188781738281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,16,1,balanced,0.8861222076416017
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,16,1,balanced,1.1643475341796876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,16,1,balanced,1.743333740234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,16,1,balanced,2.3134291076660154
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,16,1,balanced,0.03699903964996338
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,16,1,balanced,0.04941440105438232
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,16,1,balanced,0.06579135894775391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,16,1,balanced,0.06639423847198486
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,16,1,balanced,0.07511744022369385
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,16,1,balanced,0.07573760032653809
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,16,1,balanced,0.07722559928894043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,16,1,balanced,0.07799808025360108
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,16,1,balanced,0.07834752082824707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,16,1,balanced,0.07938367843627929
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,16,1,balanced,3.456195983886719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,16,1,balanced,0.08076543807983398
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,16,1,balanced,0.08580032348632813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,16,1,balanced,0.08428928375244141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,16,1,balanced,0.08942336082458496
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,16,1,balanced,0.09026880264282226
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,16,1,balanced,0.0958841609954834
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,16,1,balanced,0.09335488319396973
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,16,1,balanced,0.1133683204650879
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,16,1,balanced,0.11881792068481445
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,8,2,power_law_1.2,0.8880806732177735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,16,1,balanced,0.18223487854003906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,8,2,power_law_1.2,1.4936320495605468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,16,1,balanced,0.3121587181091309
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,16,1,balanced,0.39461952209472656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,16,1,balanced,0.6043775939941407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,16,1,balanced,4.604946594238281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,16,1,balanced,0.20811967849731444
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,16,1,power_law_1.01,0.07274176120758057
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,16,1,power_law_1.01,0.07921408176422119
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,16,1,balanced,1.1948902130126953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,16,1,power_law_1.01,0.0989408016204834
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,16,1,power_law_1.01,0.14844160079956054
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,16,1,power_law_1.01,0.06616960048675538
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,16,1,power_law_1.01,0.1457190418243408
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,16,1,power_law_1.01,0.1507532787322998
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,16,1,power_law_1.01,0.1503609561920166
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,16,1,power_law_1.01,0.1507583999633789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,16,1,balanced,1.566321258544922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,16,1,power_law_1.01,0.15923583984375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,16,1,power_law_1.01,0.15971327781677247
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,16,1,power_law_1.01,0.1607993507385254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,16,1,power_law_1.01,0.1732819175720215
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,16,1,power_law_1.01,0.18138111114501954
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,8,2,power_law_1.2,4.365710144042969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,16,1,power_law_1.01,0.18416959762573243
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,16,1,balanced,0.7687494659423828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,16,1,power_law_1.01,0.1936422348022461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,16,1,power_law_1.01,0.2064691162109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,16,1,power_law_1.01,0.29732479095458986
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,16,1,power_law_1.01,0.3866355133056641
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,16,1,power_law_1.01,0.4926950454711914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,16,1,power_law_1.01,0.6353702545166016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,16,1,power_law_1.01,0.9137709045410156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,16,1,power_law_1.01,0.041523199081420895
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,16,1,power_law_1.01,0.04928256034851074
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,16,1,power_law_1.01,0.051983361244201665
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,16,1,power_law_1.01,0.06557631969451905
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,16,1,power_law_1.01,0.07507584095001221
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,16,1,power_law_1.01,1.194263687133789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,16,1,power_law_1.01,0.0756774377822876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,16,1,power_law_1.01,0.07851456165313721
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,16,1,power_law_1.01,0.07792640209197998
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,16,1,power_law_1.01,0.07991040229797364
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,16,1,power_law_1.01,0.07927296161651612
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,16,1,power_law_1.01,0.08183808326721191
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,16,1,power_law_1.01,0.08272128105163574
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,16,1,power_law_1.01,0.0852569580078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,16,1,power_law_1.01,0.08739199638366699
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,16,1,power_law_1.01,0.09087615966796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,16,1,power_law_1.01,0.09684288024902343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,16,1,power_law_1.01,0.10073280334472656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,16,1,power_law_1.01,0.1461235237121582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,16,1,power_law_1.01,1.7664218139648438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,16,1,power_law_1.01,0.15706879615783692
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,16,1,power_law_1.01,0.1889664077758789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,16,1,power_law_1.01,0.25105215072631837
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,16,1,power_law_1.01,0.34104766845703127
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,16,1,power_law_1.01,2.3371136474609377
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,16,1,power_law_1.01,0.43285694122314455
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,8,2,power_law_1.2,6.362361450195313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,16,1,power_law_1.2,0.06267199993133546
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,16,1,power_law_1.2,0.0727129602432251
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,16,1,power_law_1.2,0.07654784202575685
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,16,1,power_law_1.2,0.0990118408203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,16,1,power_law_1.2,0.142096004486084
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,16,1,power_law_1.2,0.1461516761779785
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,16,1,power_law_1.2,0.14881664276123047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,16,1,power_law_1.2,0.14990271568298338
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,16,1,power_law_1.01,0.6220460891723633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,16,1,power_law_1.2,0.15279871940612794
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,16,1,power_law_1.2,0.15643903732299805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,16,1,power_law_1.2,0.15993727684020995
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,16,1,power_law_1.2,0.16262079238891602
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,16,1,power_law_1.2,0.16825727462768553
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,16,1,power_law_1.2,0.18157440185546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,16,1,power_law_1.2,0.18475135803222656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,16,1,power_law_1.2,0.19677631378173827
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,8,2,power_law_1.2,1.939404754638672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,16,1,power_law_1.01,3.491824645996094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,16,1,power_law_1.2,0.20338623046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,16,1,power_law_1.2,0.2952627182006836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,16,1,power_law_1.2,0.3864934539794922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,16,1,power_law_1.01,0.7854310607910155
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,16,1,power_law_1.2,0.4920614242553711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,16,1,power_law_1.2,0.6386662292480468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,16,1,power_law_1.01,1.1573772430419922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,16,1,power_law_1.2,0.9194969940185548
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,16,1,power_law_1.2,0.0413427209854126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,16,1,power_law_1.2,0.04883967876434327
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,16,1,power_law_1.2,0.0509932804107666
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,16,1,power_law_1.01,4.6402099609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,16,1,power_law_1.2,0.06619391918182374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,16,1,power_law_1.2,0.07606592178344726
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,16,1,power_law_1.2,0.07655231952667237
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,16,1,power_law_1.2,0.07839295864105225
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,16,1,power_law_1.2,0.07829760074615479
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,16,1,power_law_1.2,0.0796723222732544
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,16,1,power_law_1.2,0.0793881607055664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,16,1,power_law_1.2,0.08191103935241699
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,16,1,power_law_1.2,0.08331968307495116
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,16,1,power_law_1.2,0.08423744201660156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,16,1,power_law_1.2,0.08726528167724609
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,16,1,power_law_1.2,0.09139583587646485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,16,1,power_law_1.2,1.1966957092285155
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,16,1,power_law_1.2,0.09946240425109863
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,16,1,power_law_1.2,0.1002406406402588
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,16,1,power_law_1.2,0.14103424072265625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,16,1,power_law_1.2,0.15562239646911621
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,16,1,power_law_1.2,0.18619071960449218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,16,1,power_law_1.01,1.5716998291015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,16,1,power_law_1.2,0.25004991531372067
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,16,1,power_law_1.2,1.7653517150878906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,16,1,power_law_1.2,0.341082878112793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,16,1,power_law_1.2,0.4290387344360352
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,1,16,balanced,0.06844672203063965
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,1,16,balanced,0.06843135833740235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,1,16,balanced,0.07929215908050538
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,16,1,power_law_1.2,2.333256988525391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,1,16,balanced,0.20402816772460936
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,1,16,balanced,0.42663040161132815
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,1,16,balanced,0.10801919937133789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,1,16,balanced,0.534133758544922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,1,16,balanced,0.5343641662597657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,1,16,balanced,0.5388761520385742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,1,16,balanced,0.5385126495361329
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,1,16,balanced,0.5444569778442383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,1,16,balanced,0.5504524612426758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,1,16,balanced,0.5518963241577148
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,1,16,balanced,0.5488729476928711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,1,16,balanced,0.5593727874755859
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,1,16,balanced,0.5640947341918945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,1,16,balanced,0.577264633178711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,1,16,balanced,0.6064128112792969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,1,16,balanced,0.6423782348632813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,16,1,power_law_1.2,0.6256083297729492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,1,16,balanced,0.695726089477539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,1,16,balanced,0.7406297302246093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,1,16,balanced,0.893438720703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,1,16,balanced,1.0109107208251953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,1,16,balanced,1.41525634765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,1,16,balanced,1.7857830810546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,1,16,balanced,0.0424179220199585
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,1,16,balanced,0.041692161560058595
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,1,16,balanced,0.04939008235931396
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,1,16,balanced,0.06589183807373047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,1,16,balanced,2.6416781616210936
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,1,16,balanced,0.10812928199768065
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,1,16,balanced,0.19405311584472656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,1,16,balanced,0.21008512496948245
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,1,16,balanced,0.2100454330444336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,1,16,balanced,0.2164262390136719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,1,16,balanced,0.21620864868164064
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,1,16,balanced,3.3039129638671874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,1,16,balanced,0.21987199783325195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,1,16,balanced,0.22362367630004884
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,1,16,balanced,0.22099456787109376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,1,16,balanced,0.22496896743774414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,1,16,balanced,0.22862079620361325
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,1,16,balanced,0.23622528076171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,1,16,balanced,0.25131391525268554
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,1,16,balanced,0.26692352294921873
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,1,16,balanced,0.3017484855651856
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,1,16,balanced,0.22794752120971679
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,1,16,balanced,0.3308480072021484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,16,1,power_law_1.2,0.7892230224609376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,1,16,balanced,0.3866060638427734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,1,16,balanced,0.44957695007324217
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,1,16,balanced,0.6289011383056641
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,1,16,balanced,0.7616242980957031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,1,16,power_law_1.01,0.2005567932128906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,1,16,balanced,1.0768691253662108
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,1,16,power_law_1.01,0.4250048065185547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,1,16,power_law_1.01,0.421525764465332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,1,16,balanced,1.4034495544433594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,1,16,power_law_1.01,0.42473087310791013
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,1,16,power_law_1.01,0.42808448791503906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,1,16,power_law_1.01,0.4580748748779297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,1,16,power_law_1.01,0.49735679626464846
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,1,16,power_law_1.01,0.4259584045410156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,1,16,power_law_1.01,0.4115071868896484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,1,16,power_law_1.01,0.4219481658935547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,1,16,power_law_1.01,0.4392025756835937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,1,16,power_law_1.01,0.4509734344482422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,1,16,power_law_1.01,0.46172031402587893
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,16,1,power_law_1.2,3.480823669433594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,1,16,power_law_1.01,0.5361804962158203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,1,16,power_law_1.01,0.5452339172363281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,1,16,power_law_1.01,0.5756339263916016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,1,16,power_law_1.01,0.6377024078369141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,1,16,power_law_1.01,0.6936601257324219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,1,16,power_law_1.01,0.7309260559082031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,1,16,power_law_1.01,0.8978227233886719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,1,16,power_law_1.01,1.0622758483886718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,1,16,power_law_1.01,1.2894502258300782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,1,16,power_law_1.01,1.8169215393066405
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,1,16,power_law_1.01,0.10714879989624024
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,1,16,power_law_1.01,0.1909760093688965
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,1,16,power_law_1.01,2.305015106201172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,1,16,power_law_1.01,0.1786137580871582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,1,16,power_law_1.01,0.19112064361572265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,1,16,power_law_1.01,0.19531263351440428
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,1,16,power_law_1.01,0.16638463973999024
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,1,16,power_law_1.01,3.080979309082031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,1,16,power_law_1.01,0.2016524887084961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,1,16,power_law_1.01,0.1991462326049805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,1,16,power_law_1.01,0.1822643280029297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,1,16,power_law_1.01,0.20189695358276366
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,1,16,power_law_1.01,0.2163340759277344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,1,16,power_law_1.01,0.21179519653320314
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,1,16,power_law_1.01,0.21554176330566407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,1,16,power_law_1.01,0.21951616287231443
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,1,16,power_law_1.01,0.21264896392822266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,1,16,power_law_1.01,4.786355285644531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,1,16,power_law_1.01,0.21641727447509768
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,1,16,power_law_1.01,0.2315648078918457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,1,16,power_law_1.01,0.2676595115661621
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,1,16,power_law_1.01,0.30747520446777343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,1,16,power_law_1.01,0.34968318939208987
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,1,16,power_law_1.01,0.4177574539184571
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,1,16,power_law_1.01,0.5425497436523438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,1,16,power_law_1.01,0.6302054214477539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,1,16,power_law_1.01,0.8369165039062499
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,16,1,power_law_1.2,1.191404800415039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,1,16,power_law_1.01,6.513425903320313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,1,16,power_law_1.01,1.1993587493896485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,1,16,power_law_1.2,0.42574207305908207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,1,16,power_law_1.2,0.3537676620483398
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,1,16,power_law_1.2,0.42680446624755863
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,1,16,power_law_1.01,1.9217881774902346
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,1,16,power_law_1.2,0.4233804702758789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,1,16,power_law_1.2,0.35531646728515626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,1,16,power_law_1.2,0.20251264572143554
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,1,16,power_law_1.2,0.4669696044921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,1,16,power_law_1.01,2.181168670654297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,1,16,power_law_1.2,0.3957503890991211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,1,16,power_law_1.2,0.41705982208251957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,1,16,power_law_1.2,0.4333913421630859
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,1,16,power_law_1.2,0.45307262420654293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,1,16,power_law_1.2,0.5353497695922852
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,1,16,power_law_1.2,0.5687897491455078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,1,16,power_law_1.2,0.5336025619506837
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,1,16,power_law_1.2,0.6217587280273438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,1,16,power_law_1.2,0.6240998458862305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,1,16,power_law_1.2,0.4562470245361328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,1,16,power_law_1.2,0.664189453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,1,16,power_law_1.2,0.7938329315185546
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,1,16,power_law_1.2,1.0302732849121095
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,1,16,power_law_1.2,1.1313651275634764
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,1,16,power_law_1.2,1.6352000427246094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,1,16,power_law_1.2,0.10782336235046386
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,1,16,power_law_1.2,2.148404541015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,1,16,power_law_1.2,0.176759033203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,1,16,power_law_1.2,0.16932096481323242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,1,16,power_law_1.2,0.19229183197021485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,1,16,power_law_1.2,0.19406208038330078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,1,16,power_law_1.2,2.9204339599609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,1,16,power_law_1.2,0.1761510467529297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,1,16,power_law_1.2,0.18851327896118164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,1,16,power_law_1.2,0.1923980712890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,16,1,power_law_1.2,4.633274841308594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,1,16,power_law_1.2,0.19549055099487306
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,1,16,power_law_1.2,3.869337463378906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,1,16,power_law_1.2,0.21038719177246096
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,1,16,power_law_1.2,0.21150976181030273
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,1,16,power_law_1.2,0.21823360443115236
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,1,16,power_law_1.2,0.2145292854309082
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,1,16,power_law_1.2,0.21053951263427734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,1,16,power_law_1.2,0.2212006378173828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,1,16,power_law_1.2,0.17380352020263673
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,1,16,power_law_1.2,0.22691583633422852
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,1,16,power_law_1.2,0.29355136871337895
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,1,16,power_law_1.2,0.2985651206970215
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,1,16,power_law_1.2,0.3798566436767578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,1,16,power_law_1.2,0.4517465591430664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,1,16,power_law_1.2,0.595676155090332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,1,16,power_law_1.2,5.482034912109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,1,16,power_law_1.2,0.7097305297851563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,2,8,balanced,0.05937280178070068
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,1,16,power_law_1.2,1.0727590179443358
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,2,8,balanced,0.06579455852508545
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,2,8,balanced,0.1163263988494873
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,2,8,balanced,0.0778547191619873
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,1,16,power_law_1.2,1.3810009765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,2,8,balanced,0.19136768341064453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,2,8,balanced,0.4142361450195312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,2,8,balanced,0.5293964767456055
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,2,8,balanced,0.5298291015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,2,8,balanced,0.5351385498046876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,1,16,power_law_1.2,2.336565704345703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,2,8,balanced,0.5399283218383789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,2,8,balanced,0.5380390548706055
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,2,8,balanced,0.5438655853271485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,2,8,balanced,0.5511500930786133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,2,8,balanced,0.5561100769042968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,2,8,balanced,0.5665292739868164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,1,16,power_law_1.2,3.2444747924804687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,2,8,balanced,0.5737011337280273
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,2,8,balanced,0.5932659149169922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,2,8,balanced,0.6326425552368165
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,1,16,power_law_1.2,9.002321777343749
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,2,8,balanced,0.6821478271484376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,2,8,balanced,0.7588365173339844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,2,8,balanced,0.8279039764404297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,2,8,balanced,0.03375999927520752
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,2,8,balanced,1.0081356811523439
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,2,8,balanced,0.040293121337890626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,2,8,balanced,1.1703424072265625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,2,8,balanced,0.04955647945404053
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,2,8,balanced,0.0684761619567871
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,16,1,power_law_1.2,1.5870016479492188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,2,8,balanced,0.19612159729003906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,2,8,balanced,1.647187194824219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,2,8,balanced,0.21804927825927733
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,2,8,balanced,0.22318208694458005
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,2,8,balanced,0.22249471664428713
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,2,8,balanced,0.22779264450073242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,2,8,balanced,0.21703680038452147
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,2,8,balanced,2.0833740234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,2,8,balanced,0.22199296951293945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,2,8,balanced,0.22289152145385743
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,2,8,balanced,0.23127552032470705
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,2,8,balanced,0.2369024085998535
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,2,8,balanced,0.26160512924194335
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,2,8,balanced,0.11269375801086426
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,2,8,balanced,0.2856934356689453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,2,8,balanced,0.31254144668579104
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,2,8,balanced,0.357891845703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,2,8,balanced,0.4044659042358399
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,2,8,balanced,3.0667724609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,2,8,balanced,0.5075571060180664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,2,8,balanced,0.6179174423217774
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,2,8,power_law_1.01,0.11622143745422364
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,2,8,balanced,0.24908031463623045
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,2,8,balanced,0.8769638061523437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,2,8,power_law_1.01,0.1844428825378418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,2,8,power_law_1.01,0.23480703353881838
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,2,8,balanced,1.0786879730224608
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,2,8,balanced,3.8880487060546876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,2,8,power_law_1.01,0.27912063598632814
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,2,8,power_law_1.01,0.3081587219238281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,2,8,power_law_1.01,0.37713279724121096
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,2,8,power_law_1.01,0.4416486358642578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,2,8,power_law_1.01,0.41392639160156247
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,2,8,balanced,1.5605503845214845
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,2,8,power_law_1.01,0.44778881072998045
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,2,8,power_law_1.01,0.4296550369262695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,2,8,power_law_1.01,0.46462078094482423
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,2,8,power_law_1.01,0.46613502502441406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,2,8,power_law_1.01,0.46866432189941404
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,2,8,balanced,2.0465061950683596
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,2,8,power_law_1.01,0.4954470443725586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,2,8,power_law_1.01,0.5041228866577148
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,2,8,power_law_1.01,0.5425625610351562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,2,8,power_law_1.01,0.5508505630493163
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,2,8,power_law_1.01,0.666138916015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,2,8,power_law_1.01,0.7080934143066406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,2,8,power_law_1.01,0.8177855682373046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,2,8,power_law_1.01,0.8910208129882813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,2,8,power_law_1.01,0.06764927864074707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,2,8,power_law_1.01,1.2167398071289062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,2,8,power_law_1.01,0.10941311836242676
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,2,8,power_law_1.01,1.4076748657226563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,2,8,power_law_1.01,0.12508928298950195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,2,8,power_law_1.01,0.15270527839660644
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,2,8,power_law_1.01,0.16268928527832033
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,2,8,power_law_1.01,0.185350399017334
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,2,8,power_law_1.01,0.18687871932983396
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,2,8,power_law_1.01,0.19016704559326172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,2,8,power_law_1.01,2.0465242004394533
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,2,8,power_law_1.01,0.19749631881713867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,2,8,power_law_1.01,0.1957043266296387
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,2,8,power_law_1.01,0.20397056579589842
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,2,8,power_law_1.01,0.2019417572021484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,2,8,power_law_1.01,0.20046976089477536
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,2,8,power_law_1.01,2.533032989501953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,2,8,power_law_1.01,0.2058086395263672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,2,8,power_law_1.01,0.20996095657348635
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,2,8,power_law_1.01,0.21404031753540037
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,2,8,power_law_1.01,0.22501888275146484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,2,8,power_law_1.01,0.2640985679626465
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,2,8,power_law_1.01,0.31041023254394534
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,2,8,power_law_1.01,0.3604288101196289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,2,8,power_law_1.01,0.39855873107910156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,2,8,power_law_1.01,0.5385382461547852
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,2,8,power_law_1.01,0.6499314880371093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,2,8,power_law_1.2,0.11659135818481445
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,2,8,power_law_1.01,4.339566040039062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,2,8,power_law_1.01,1.290645751953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,2,8,power_law_1.2,0.18400768280029295
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,2,8,power_law_1.2,0.18215040206909178
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,2,8,power_law_1.2,0.2415116882324219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,2,8,power_law_1.2,0.36064510345458983
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,2,8,power_law_1.01,1.6850572204589844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,2,8,power_law_1.01,5.2209973144531245
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,2,8,power_law_1.01,0.9512576293945312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,2,8,power_law_1.2,0.35986049652099605
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,2,8,power_law_1.2,0.39593856811523437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,2,8,power_law_1.2,0.4101055908203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,2,8,power_law_1.2,0.4124095916748047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,2,8,power_law_1.01,2.200806427001953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,2,8,power_law_1.2,0.4420633697509766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,2,8,power_law_1.2,0.4353855895996094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,2,8,power_law_1.2,0.46174976348876956
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,2,8,power_law_1.2,0.45993728637695314
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,2,8,power_law_1.2,0.5026047897338868
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,2,8,power_law_1.2,0.5129740905761719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,2,8,power_law_1.2,0.5979520034790039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,2,8,power_law_1.2,0.6915570831298827
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,2,8,power_law_1.2,0.7136627197265625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,2,8,power_law_1.2,0.8604338836669921
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,2,8,power_law_1.2,0.9688780975341796
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,2,8,power_law_1.2,0.0666815996170044
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,2,8,power_law_1.2,1.3176077270507813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,2,8,power_law_1.2,0.10369024276733399
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,2,8,power_law_1.2,0.6643878173828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,2,8,power_law_1.2,0.11277567863464355
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,2,8,power_law_1.2,0.14669440269470216
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,2,8,power_law_1.2,0.1754060745239258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,2,8,power_law_1.2,0.17725055694580077
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,2,8,power_law_1.2,2.0302476501464843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,2,8,power_law_1.2,0.18333568572998046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,2,8,power_law_1.2,0.1897792053222656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,2,8,power_law_1.2,0.19099391937255858
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,2,8,power_law_1.2,0.2004684829711914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,2,8,power_law_1.2,0.2055308723449707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,2,8,power_law_1.2,1.6932403564453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,2,8,power_law_1.2,0.20610687255859378
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,2,8,power_law_1.2,0.19280895233154297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,2,8,power_law_1.2,0.2068953514099121
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,2,8,power_law_1.2,0.20805376052856445
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,2,8,power_law_1.2,0.2191168022155762
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,2,8,power_law_1.2,0.2435852813720703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,2,8,power_law_1.2,0.281265926361084
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,2,8,power_law_1.2,0.3227571105957031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,2,8,power_law_1.2,0.3687795257568359
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,2,8,power_law_1.2,0.4292044830322266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,2,8,power_law_1.2,0.5599987030029296
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,2,8,power_law_1.2,0.688006362915039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,2,8,power_law_1.2,2.8720269775390626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,4,4,balanced,0.05537663936614991
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,4,4,balanced,0.06588160037994385
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,2,8,power_law_1.2,1.0209919738769533
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,4,4,balanced,0.07926911830902099
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,4,4,balanced,0.11124863624572753
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,4,4,balanced,0.1821171188354492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,2,8,power_law_1.2,1.3569293212890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,4,4,balanced,0.40798206329345704
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,4,4,balanced,0.4113471984863281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,4,4,balanced,0.4138956832885742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,4,4,balanced,0.5128153610229492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,2,8,power_law_1.2,1.7901568603515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,4,4,balanced,0.5122623825073243
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,4,4,balanced,0.5182745742797852
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,4,4,balanced,0.5242854309082031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,4,4,balanced,0.5293939208984375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,4,4,balanced,0.5376844787597657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,4,4,balanced,0.5504819107055664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,4,4,balanced,0.561786880493164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,4,4,balanced,0.5723366546630859
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,2,8,power_law_1.2,4.169591064453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,2,8,power_law_1.2,2.716698913574219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,4,4,balanced,0.6085248184204102
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,4,4,balanced,0.6503449249267578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,4,4,balanced,0.7303513336181641
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,4,4,balanced,0.8096511840820313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,2,8,power_law_1.2,6.178698120117188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,4,4,balanced,0.03183232069015503
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,4,4,balanced,1.0146867370605468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,4,4,balanced,0.04143104076385498
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,4,4,balanced,0.05111423969268799
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,4,4,balanced,1.1871449279785158
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,4,4,balanced,0.07206912040710449
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,4,4,balanced,0.11589376449584961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,4,4,balanced,0.20218879699707032
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,4,4,balanced,0.20792448043823245
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,4,4,balanced,0.21034624099731447
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,4,4,balanced,0.23272192001342776
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,4,4,balanced,0.2336934471130371
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,4,4,balanced,0.23651968002319337
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,4,4,balanced,0.2389811134338379
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,4,4,balanced,2.0560231018066406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,4,4,balanced,0.24128768920898436
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,4,4,balanced,0.24754175186157226
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,4,4,balanced,0.2550041580200195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,4,4,balanced,0.26020864486694334
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,4,4,balanced,0.27100543975830077
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,4,4,balanced,0.29728256225585936
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,4,4,balanced,0.3179852867126465
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,4,4,balanced,0.36474624633789066
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,4,4,balanced,0.4091379165649414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,4,4,balanced,3.001758728027344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,4,4,balanced,0.4980812835693359
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,4,4,balanced,0.5977203369140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,4,4,power_law_1.01,0.07958784103393554
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,4,4,balanced,0.8761869049072265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,4,4,balanced,1.640716857910156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,4,4,balanced,3.8534848022460935
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,4,4,power_law_1.01,0.10845696449279785
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,4,4,balanced,1.057791976928711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,4,4,power_law_1.01,0.12592127799987793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,4,4,power_law_1.01,0.16573440551757812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,4,4,power_law_1.01,0.2075596809387207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,4,4,power_law_1.01,0.21864831924438474
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,4,4,balanced,1.5395277404785157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,4,4,power_law_1.01,0.3455744171142578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,4,4,power_law_1.01,0.3008870315551758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,4,4,power_law_1.01,0.4069977569580078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,4,4,power_law_1.01,0.412250862121582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,4,4,power_law_1.01,0.43011199951171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,4,4,power_law_1.01,0.45830398559570307
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,4,4,balanced,2.00732421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,4,4,power_law_1.01,0.46998142242431645
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,4,4,power_law_1.01,0.4758374404907227
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,4,4,power_law_1.01,0.5048371124267578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,4,4,power_law_1.01,0.5309132766723633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,4,4,power_law_1.01,0.6670028686523437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,4,4,power_law_1.01,0.6831667327880859
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,4,4,power_law_1.01,0.7323519897460937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,4,4,power_law_1.01,0.861866226196289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,4,4,power_law_1.01,0.4480908966064453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,4,4,power_law_1.01,1.1600729370117189
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,4,4,power_law_1.01,0.07031680107116699
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,4,4,power_law_1.01,1.3588710021972656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,4,4,power_law_1.01,0.08920319557189942
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,4,4,power_law_1.01,0.11073792457580567
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,4,4,power_law_1.01,0.14005120277404787
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,4,4,power_law_1.01,0.15524991989135742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,4,4,power_law_1.01,1.9510873413085938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,4,4,power_law_1.01,0.1805196762084961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,4,4,power_law_1.01,0.05065343856811524
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,4,4,power_law_1.01,0.18603776931762694
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,4,4,power_law_1.01,0.20527360916137694
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,4,4,power_law_1.01,0.20487295150756837
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,4,4,power_law_1.01,0.21627136230468752
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,4,4,power_law_1.01,2.3330700683593752
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,4,4,power_law_1.01,0.21180032730102538
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,4,4,power_law_1.01,0.22197759628295896
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,4,4,power_law_1.01,0.22907520294189454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,4,4,power_law_1.01,0.23785472869873048
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,4,4,power_law_1.01,0.24712575912475584
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,4,4,power_law_1.01,0.2867430305480957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,4,4,power_law_1.01,0.3203046417236328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,4,4,power_law_1.01,0.3820275115966797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,4,4,power_law_1.01,0.4314598464965821
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,4,4,power_law_1.01,3.417576904296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,4,4,power_law_1.01,0.20995967864990234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,4,4,power_law_1.01,0.5657459259033203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,4,4,power_law_1.01,0.6781222534179687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,4,4,power_law_1.2,0.07930751800537109
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,4,4,power_law_1.2,0.10385279655456543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,4,4,power_law_1.01,0.9626777648925782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,4,4,power_law_1.2,0.16627584457397462
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,4,4,power_law_1.2,0.19848703384399413
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,4,4,power_law_1.01,1.2196288299560547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,4,4,power_law_1.2,0.2502707290649414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,4,4,power_law_1.01,4.623960266113281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,4,4,power_law_1.2,0.2919385528564453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,4,4,power_law_1.2,0.2979148864746094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,4,4,power_law_1.01,1.6515840148925782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,4,4,power_law_1.2,0.4361228942871094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,4,4,power_law_1.2,0.1158566379547119
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,4,4,power_law_1.2,0.4299033737182617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,4,4,power_law_1.2,0.45678974151611323
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,4,4,power_law_1.2,0.4684979248046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,4,4,power_law_1.01,2.243616027832031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,4,4,power_law_1.2,0.5029375839233399
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,4,4,power_law_1.2,0.484851188659668
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,4,4,power_law_1.2,0.5487449645996094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,4,4,power_law_1.2,0.6078400039672852
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,4,4,power_law_1.2,0.6852710723876954
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,4,4,power_law_1.2,0.6945113372802735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,4,4,power_law_1.2,0.4109759902954101
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,4,4,power_law_1.2,0.7821465301513671
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,4,4,power_law_1.2,0.8987776184082031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,4,4,power_law_1.2,0.050680317878723145
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,4,4,power_law_1.2,1.19889404296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,4,4,power_law_1.2,0.07020544052124024
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,4,4,power_law_1.2,0.07574783802032471
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,4,4,power_law_1.2,1.4544781494140626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,4,4,power_law_1.2,0.1062502384185791
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,4,4,power_law_1.2,0.11682559967041015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,4,4,power_law_1.2,0.13816191673278808
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,4,4,power_law_1.2,1.9963046264648436
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,4,4,power_law_1.2,0.18499711990356443
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,4,4,power_law_1.2,0.20461824417114255
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,4,4,power_law_1.2,0.2110335922241211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,4,4,power_law_1.2,2.715407409667969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,4,4,power_law_1.2,0.21929216384887695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,4,4,power_law_1.2,0.17650175094604492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,4,4,power_law_1.2,0.2178828811645508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,4,4,power_law_1.2,0.22895360946655274
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,4,4,power_law_1.2,0.23181823730468748
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,4,4,power_law_1.2,0.21101184844970705
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,4,4,power_law_1.2,0.2462924766540527
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,4,4,power_law_1.2,0.2642278480529785
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,4,4,power_law_1.2,0.31041280746459965
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,4,4,power_law_1.2,3.7068249511718747
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,4,4,power_law_1.2,0.3451289749145508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,4,4,power_law_1.2,0.3959603118896484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,4,4,power_law_1.2,0.4552691268920898
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,4,4,power_law_1.2,0.58548095703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,8,2,balanced,0.055599360466003424
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,8,2,balanced,0.06437119960784912
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,4,4,power_law_1.2,0.6815142059326172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,8,2,balanced,0.07707520008087158
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,8,2,balanced,0.10863488197326661
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,4,4,power_law_1.2,4.797173767089843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,8,2,balanced,0.18829952239990236
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,4,4,power_law_1.2,0.9650508880615234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,8,2,balanced,0.41014785766601564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,8,2,balanced,0.41019134521484374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,8,2,balanced,0.4155686569213867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,8,2,balanced,0.4195417785644532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,8,2,balanced,0.4214668655395508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,8,2,balanced,0.5126553726196289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,4,4,power_law_1.2,1.7104856872558591
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,8,2,balanced,0.517446403503418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,8,2,balanced,0.5261798477172851
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,8,2,balanced,0.5341401672363281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,4,4,power_law_1.2,1.2644633483886718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,8,2,balanced,0.5402201461791992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,8,2,balanced,0.5540019226074219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,8,2,balanced,0.41296127319335935
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,4,4,power_law_1.2,2.387535400390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,8,2,balanced,0.6282137680053711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,8,2,balanced,0.7018879699707031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,8,2,balanced,0.7801395416259765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,8,2,balanced,1.0423987579345702
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,8,2,balanced,0.03227776050567627
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,8,2,balanced,0.04181248188018799
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,8,2,balanced,0.05497216224670411
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,8,2,balanced,1.2382195281982422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,8,2,balanced,0.07785088062286377
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,8,2,balanced,0.13255295753479004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,8,2,balanced,0.23929344177246095
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,8,2,balanced,1.6887437438964845
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,8,2,balanced,0.23933055877685544
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,8,2,balanced,0.23999488830566404
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,8,2,balanced,0.24320640563964843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,8,2,balanced,0.24594688415527344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,8,2,balanced,2.1029029846191407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,8,2,balanced,0.24781824111938477
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,8,2,balanced,0.2916543960571289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,8,2,balanced,0.5907404708862305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,8,2,balanced,0.29621631622314454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,8,2,balanced,0.3007116889953613
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,8,2,balanced,0.31167360305786135
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,8,2,balanced,0.30956544876098635
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,8,2,balanced,0.317262077331543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,8,2,balanced,0.3342668914794922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,8,2,balanced,0.35709438323974607
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,8,2,balanced,0.39718654632568356
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,8,2,balanced,3.0222924804687503
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,8,2,balanced,0.4419327926635742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,8,2,balanced,0.5256959915161132
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,8,2,power_law_1.01,0.06398975849151611
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,8,2,balanced,0.6138227081298828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,8,2,power_law_1.01,0.0769267177581787
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,8,2,power_law_1.01,0.09311615943908691
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,8,2,balanced,0.9414733123779296
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,8,2,power_law_1.01,0.12415231704711913
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,8,2,power_law_1.01,0.16646272659301758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,8,2,balanced,3.9021682739257812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,8,2,balanced,1.122740478515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,8,2,power_law_1.01,0.1958233642578125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,8,2,power_law_1.01,0.2788723182678223
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,8,2,power_law_1.01,0.28816640853881836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,8,2,power_law_1.01,0.3127039909362793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,8,2,power_law_1.01,0.3365529632568359
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,8,2,balanced,1.6255168151855468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,8,2,power_law_1.01,0.39230976104736326
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,8,2,power_law_1.01,0.4235289764404297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,8,2,power_law_1.01,0.42124286651611326
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,8,2,power_law_1.01,0.440975341796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,8,2,power_law_1.01,0.44207870483398437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,8,2,balanced,2.1264498901367186
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,8,2,power_law_1.01,0.47022846221923825
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,8,2,power_law_1.01,0.5178073501586914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,8,2,power_law_1.01,0.6536204528808593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,8,2,power_law_1.01,0.6845030212402344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,8,2,power_law_1.01,0.7318233489990235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,8,2,power_law_1.01,0.8250790405273438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,8,2,power_law_1.01,1.0575065612792969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,8,2,power_law_1.01,1.2773158264160156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,8,2,power_law_1.01,0.05459455966949463
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,8,2,power_law_1.01,0.06642303943634034
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,8,2,power_law_1.01,0.09061120033264161
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,8,2,power_law_1.01,1.7607948303222656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,8,2,power_law_1.01,0.11793279647827148
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,8,2,power_law_1.01,0.041300477981567385
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,8,2,power_law_1.01,0.15070464134216308
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,8,2,power_law_1.01,0.18799232482910155
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,8,2,power_law_1.01,2.2152755737304686
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,8,2,power_law_1.01,0.19603584289550782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,8,2,power_law_1.01,0.2036697578430176
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,8,2,power_law_1.01,0.2089894485473633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,8,2,power_law_1.01,0.21930879592895508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,8,2,power_law_1.01,0.2555839920043945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,8,2,power_law_1.01,0.2590988731384277
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,8,2,power_law_1.01,0.26508928298950196
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,8,2,power_law_1.01,0.27168256759643555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,8,2,power_law_1.01,0.2788006401062012
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,8,2,power_law_1.01,0.2907993507385254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,8,2,power_law_1.01,0.3155891227722168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,8,2,power_law_1.01,3.1337701416015626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,8,2,power_law_1.01,0.35657470703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,8,2,power_law_1.01,0.43626880645751953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,8,2,power_law_1.01,0.48123775482177733
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,8,2,power_law_1.01,0.6116966247558594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,8,2,power_law_1.2,0.06382207870483399
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,8,2,power_law_1.01,0.7334822082519532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,8,2,power_law_1.2,0.07559296131134033
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,8,2,power_law_1.2,0.08670847892761231
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,8,2,power_law_1.01,4.070530700683594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,8,2,power_law_1.2,0.1181926441192627
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,8,2,power_law_1.2,0.15119104385375975
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,8,2,power_law_1.2,0.1730291175842285
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,8,2,power_law_1.01,1.256505584716797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,8,2,power_law_1.2,0.2581337547302246
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,8,2,power_law_1.2,0.27799808502197265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,8,2,power_law_1.2,0.29798912048339843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,8,2,power_law_1.2,0.3303859329223633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,8,2,power_law_1.01,1.7576396179199218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,8,2,power_law_1.2,0.3897983932495117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,8,2,power_law_1.2,0.4175526428222656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,8,2,power_law_1.2,0.42425086975097653
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,8,2,power_law_1.01,0.9943167877197265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,8,2,power_law_1.2,0.47361278533935547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,8,2,power_law_1.01,2.308702697753906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,8,2,power_law_1.2,0.5231667327880859
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,8,2,power_law_1.2,0.5613734436035156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,8,2,power_law_1.2,0.6768691253662109
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,8,2,power_law_1.2,0.6922969818115234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,8,2,power_law_1.2,0.742534408569336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,8,2,power_law_1.2,0.8364236450195313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,8,2,power_law_1.2,0.45371265411376954
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,8,2,power_law_1.2,0.04166656017303467
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,8,2,power_law_1.2,1.098819808959961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,8,2,power_law_1.2,0.05431680202484131
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,8,2,power_law_1.2,1.343504638671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,8,2,power_law_1.2,0.06314112186431885
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,8,2,power_law_1.2,0.08609791755676269
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,8,2,power_law_1.2,0.11003520011901854
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,8,2,power_law_1.2,0.12417920112609863
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,8,2,power_law_1.2,1.8450175476074218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,8,2,power_law_1.2,0.19555456161499024
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,8,2,power_law_1.2,0.20217216491699216
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,8,2,power_law_1.2,0.2045145606994629
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,8,2,power_law_1.2,0.21524991989135742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,8,2,power_law_1.2,2.2240818786621093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,8,2,power_law_1.2,0.2535411262512207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,8,2,power_law_1.2,0.2553715133666992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,8,2,power_law_1.2,0.2688294410705566
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,8,2,power_law_1.2,0.28330623626708984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,8,2,power_law_1.2,0.28829696655273435
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,8,2,power_law_1.2,0.3034316825866699
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,8,2,power_law_1.2,0.1831808090209961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,8,2,power_law_1.2,0.3304499053955078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,8,2,power_law_1.2,0.39342464447021486
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,8,2,power_law_1.2,0.4464396667480469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,8,2,power_law_1.2,0.5012531280517578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,8,2,power_law_1.2,0.6528793334960937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,8,2,power_law_1.2,0.7463629150390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,8,2,power_law_1.2,1.0238003540039062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,16,1,balanced,0.07325119972229004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,8,2,power_law_1.2,4.0975820922851565
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,16,1,balanced,0.08827072143554687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,16,1,balanced,0.1227462387084961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,8,2,power_law_1.2,1.3127500915527344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,16,1,balanced,0.20404544830322266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,16,1,balanced,0.06359360218048096
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,8,2,power_law_1.2,1.8003520202636718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,16,1,balanced,0.45027072906494137
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,16,1,balanced,0.45181121826171877
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,8,2,power_law_1.2,3.137171325683594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,16,1,balanced,0.4549568176269531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,8,2,power_law_1.2,2.3318080139160156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,16,1,balanced,0.4612204742431641
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,16,1,balanced,0.4711942291259765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,16,1,balanced,0.44324287414550784
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,16,1,balanced,0.48114879608154293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,16,1,balanced,0.5541926574707031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,16,1,balanced,0.5547795104980469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,16,1,balanced,0.5706662368774414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,16,1,balanced,0.6013644790649414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,16,1,balanced,0.46588287353515623
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,16,1,balanced,0.6352217483520508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,16,1,balanced,0.7055993652343749
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,16,1,balanced,0.4462553787231445
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,16,1,balanced,0.7837760162353515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,16,1,balanced,0.042800002098083496
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,16,1,balanced,0.055462398529052735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,16,1,balanced,1.1342444610595703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,16,1,balanced,1.1697344207763671
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,16,1,balanced,0.07189760208129883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,16,1,balanced,0.104901762008667
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,16,1,balanced,0.17563840866088867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,16,1,balanced,0.3189113616943359
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,16,1,balanced,0.32188030242919924
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,16,1,balanced,0.3233484649658203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,16,1,balanced,0.3253868865966797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,16,1,balanced,2.0588934326171873
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,16,1,balanced,0.32763519287109377
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,16,1,balanced,0.33184703826904294
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,16,1,balanced,0.33307838439941406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,16,1,balanced,0.33565311431884765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,16,1,balanced,0.34401985168457033
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,16,1,balanced,1.6351565551757812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,16,1,balanced,0.4234732818603516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,16,1,balanced,0.4265401458740234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,16,1,balanced,0.44360702514648437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,16,1,balanced,0.4630796813964844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,16,1,balanced,0.5005158233642578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,16,1,balanced,0.5353926467895508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,16,1,balanced,0.4204460906982422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,16,1,balanced,0.608092155456543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,16,1,balanced,0.6855635070800782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,16,1,power_law_1.01,0.06284287929534912
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,16,1,balanced,3.0269265747070313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,16,1,power_law_1.01,0.07333055973052979
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,16,1,balanced,4.014437866210938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,16,1,balanced,1.1010963439941406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,16,1,power_law_1.01,0.08781184196472168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,16,1,balanced,1.2617440032958984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,16,1,power_law_1.01,0.12272768020629883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,16,1,power_law_1.01,0.16682111740112304
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,16,1,power_law_1.01,0.23116607666015626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,16,1,power_law_1.01,0.30226367950439453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,16,1,power_law_1.01,0.3156723213195801
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,16,1,power_law_1.01,0.32751678466796874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,16,1,power_law_1.01,0.34630783081054684
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,16,1,power_law_1.01,0.4043008041381836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,16,1,balanced,1.8433255004882814
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,16,1,power_law_1.01,0.43252799987792967
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,16,1,power_law_1.01,0.43329406738281245
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,16,1,power_law_1.01,0.4328166580200195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,16,1,power_law_1.01,0.48420158386230466
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,16,1,power_law_1.01,0.49788864135742184
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,16,1,power_law_1.01,0.5424217605590821
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,16,1,balanced,2.436057586669922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,16,1,power_law_1.01,0.6540659332275391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,16,1,power_law_1.01,0.7122771453857422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,16,1,power_law_1.01,0.7559724426269531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,16,1,power_law_1.01,0.8301792144775391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,16,1,power_law_1.01,0.04206399917602539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,16,1,power_law_1.01,0.05500480175018311
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,16,1,power_law_1.01,1.0637471771240234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,16,1,power_law_1.01,0.07277632236480713
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,16,1,power_law_1.01,1.2692793273925782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,16,1,power_law_1.01,0.104204158782959
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,16,1,power_law_1.01,0.14569408416748048
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,16,1,power_law_1.01,0.1925484848022461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,16,1,power_law_1.01,0.25272192001342775
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,16,1,power_law_1.01,0.26134975433349605
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,16,1,power_law_1.01,1.770438995361328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,16,1,power_law_1.01,0.2673855972290039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,16,1,power_law_1.01,0.2697561645507812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,16,1,power_law_1.01,0.2837062454223633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,16,1,power_law_1.01,0.28995136260986326
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,16,1,power_law_1.01,0.2972915267944336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,16,1,power_law_1.01,2.242262420654297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,16,1,power_law_1.01,0.3021023941040039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,16,1,power_law_1.01,0.3777452850341797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,16,1,power_law_1.01,0.38342655181884766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,16,1,power_law_1.01,0.3957913589477539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,16,1,power_law_1.01,0.4258886337280273
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,16,1,power_law_1.01,0.4662092971801758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,16,1,power_law_1.01,0.5494854354858398
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,16,1,power_law_1.01,3.1648440551757813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,16,1,power_law_1.01,0.6004684829711915
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,16,1,power_law_1.01,0.7283980560302734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,16,1,power_law_1.2,0.06256896018981933
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,16,1,power_law_1.01,0.8791744232177734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,16,1,power_law_1.2,0.073470721244812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,16,1,power_law_1.2,0.08845888137817383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,16,1,power_law_1.01,1.172867202758789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,16,1,power_law_1.2,0.11629823684692382
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,16,1,power_law_1.01,4.119165954589844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,16,1,power_law_1.2,0.15924032211303712
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,16,1,power_law_1.2,0.22215103149414062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,16,1,power_law_1.01,1.4641938781738282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,16,1,power_law_1.2,0.2993260765075684
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,16,1,power_law_1.2,0.31344703674316404
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,16,1,power_law_1.2,0.3203174209594727
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,16,1,power_law_1.2,0.3586899185180664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,16,1,power_law_1.2,0.4108902359008789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,16,1,power_law_1.01,2.037151947021484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,16,1,power_law_1.2,0.42993793487548826
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,16,1,power_law_1.2,0.4330995178222656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,16,1,power_law_1.2,0.4413766479492187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,16,1,power_law_1.2,0.5015353775024414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,16,1,power_law_1.2,0.519752311706543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,16,1,power_law_1.2,0.5643001556396484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,16,1,power_law_1.01,2.62893310546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,16,1,power_law_1.2,0.6907270050048828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,16,1,power_law_1.2,0.7228729248046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,16,1,power_law_1.2,0.780176010131836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,16,1,power_law_1.2,0.8556018829345703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,16,1,power_law_1.2,0.0554201602935791
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,16,1,power_law_1.2,1.3132063293457032
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,16,1,power_law_1.2,0.07261248111724854
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,16,1,power_law_1.2,0.042526078224182126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,16,1,power_law_1.2,0.09919487953186035
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,16,1,power_law_1.2,1.775525817871094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,16,1,power_law_1.2,0.13796544075012207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,16,1,power_law_1.2,1.0858879852294923
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,16,1,power_law_1.2,0.1795564842224121
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,16,1,power_law_1.2,0.24871999740600587
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,16,1,power_law_1.2,2.277374725341797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,16,1,power_law_1.2,0.2714425659179688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,16,1,power_law_1.2,0.2808198356628418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,16,1,power_law_1.2,0.2894380760192871
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,16,1,power_law_1.2,0.29891775131225584
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,16,1,power_law_1.2,0.3071507263183594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,16,1,power_law_1.2,0.3845977783203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,16,1,power_law_1.2,0.2592025566101074
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,16,1,power_law_1.2,3.2144262695312498
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,16,1,power_law_1.2,0.4074668884277344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,16,1,power_law_1.2,0.44719039916992187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,16,1,power_law_1.2,0.503704948425293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,16,1,power_law_1.2,0.5688505554199219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,16,1,power_law_1.2,0.39369792938232423
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,16,1,power_law_1.2,4.162705993652343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,16,1,power_law_1.2,0.906670684814453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,1,16,balanced,0.0625113582611084
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,1,16,balanced,0.0624729585647583
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,16,1,power_law_1.2,1.181604461669922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,16,1,power_law_1.2,0.2627923202514649
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,1,16,balanced,0.07134335994720459
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,1,16,balanced,0.09713664054870605
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,1,16,balanced,0.15924863815307616
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,1,16,balanced,0.35005950927734375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,16,1,power_law_1.2,1.4729875183105468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,1,16,balanced,0.4664012908935547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,1,16,balanced,0.4704345703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,1,16,balanced,0.4732403182983399
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,1,16,balanced,0.4719910430908203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,16,1,power_law_1.2,0.6204339218139648
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,1,16,balanced,0.4799039840698242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,1,16,balanced,0.4850035095214844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,16,1,power_law_1.2,0.7600505828857422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,1,16,balanced,0.4839257431030274
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,16,1,power_law_1.2,2.0594732666015623
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,1,16,balanced,0.4777983856201172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,1,16,balanced,0.48999168395996096
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,1,16,balanced,0.508012809753418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,1,16,balanced,0.5374412918090821
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,1,16,balanced,0.562848014831543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,1,16,balanced,0.6063513565063476
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,16,1,power_law_1.2,2.6530075073242188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,1,16,balanced,0.6431858825683594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,1,16,balanced,0.7838694763183593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,1,16,balanced,0.03709311962127686
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,1,16,balanced,0.8843328094482421
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,1,16,balanced,0.03744127988815308
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,1,16,balanced,0.4968742370605469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,1,16,balanced,0.04569983959197998
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,1,16,balanced,0.06052864074707032
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,1,16,balanced,0.09689855575561523
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,1,16,balanced,1.2236582183837892
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,1,16,balanced,0.16857856750488281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,1,16,balanced,0.1829836845397949
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,1,16,balanced,0.18803455352783202
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,1,16,balanced,0.18944128036499025
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,1,16,balanced,0.1914406394958496
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,1,16,balanced,1.5344320678710939
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,1,16,balanced,0.19689983367919922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,1,16,balanced,0.2004364776611328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,1,16,balanced,0.19435647964477537
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,1,16,balanced,0.1994367980957031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,1,16,balanced,0.20373888015747071
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,1,16,balanced,0.20888959884643557
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,1,16,balanced,0.2241036796569824
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,1,16,balanced,0.23871999740600586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,1,16,balanced,0.2709324836730957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,1,16,balanced,0.29622783660888674
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,1,16,balanced,0.34734848022460935
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,1,16,balanced,2.2931341552734374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,1,16,balanced,0.19358463287353517
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,1,16,balanced,0.5620377731323243
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,1,16,balanced,0.6794035339355469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,1,16,power_law_1.01,0.15711359977722167
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,1,16,balanced,2.86706298828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,1,16,power_law_1.01,0.3458252716064453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,1,16,balanced,0.960469741821289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,1,16,power_law_1.01,0.3459097671508789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,1,16,power_law_1.01,0.34651775360107423
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,1,16,power_law_1.01,0.29797119140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,1,16,balanced,1.2542630767822265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,1,16,power_law_1.01,0.40052352905273436
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,1,16,power_law_1.01,0.3315033721923828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,1,16,power_law_1.01,0.34468353271484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,1,16,power_law_1.01,0.34969726562500003
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,1,16,power_law_1.01,0.4359231948852539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,1,16,power_law_1.01,0.4095526504516601
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,1,16,power_law_1.01,0.42990974426269535
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,1,16,power_law_1.01,0.4503615951538086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,1,16,balanced,0.40384769439697266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,1,16,power_law_1.01,0.3672652816772461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,1,16,power_law_1.01,0.47544063568115236
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,1,16,power_law_1.01,0.5208089447021484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,1,16,power_law_1.01,0.5998527908325195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,1,16,power_law_1.01,0.6037760162353516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,1,16,power_law_1.01,0.39934974670410156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,1,16,power_law_1.01,0.7176038360595702
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,1,16,power_law_1.01,0.8879654693603516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,1,16,power_law_1.01,0.09507712364196777
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,1,16,power_law_1.01,1.0959795379638673
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,1,16,power_law_1.01,0.16610048294067384
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,1,16,power_law_1.01,0.1647974395751953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,1,16,power_law_1.01,0.16773248672485352
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,1,16,power_law_1.01,0.16731775283813477
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,1,16,power_law_1.01,0.1590003204345703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,1,16,power_law_1.01,1.9434303283691405
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,1,16,power_law_1.01,0.16883968353271483
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,1,16,power_law_1.01,0.16524288177490234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,1,16,power_law_1.01,0.16960767745971678
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,1,16,power_law_1.01,0.16283903121948243
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,1,16,power_law_1.01,0.1796441650390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,1,16,power_law_1.01,1.4031309509277343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,1,16,power_law_1.01,0.18710271835327147
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,1,16,power_law_1.01,2.556976623535156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,1,16,power_law_1.01,0.18132991790771485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,1,16,power_law_1.01,0.18747648239135742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,1,16,power_law_1.01,0.24371072769165042
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,1,16,power_law_1.01,0.25969280242919923
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,1,16,power_law_1.01,0.2940838432312012
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,1,16,power_law_1.01,0.3546265411376953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,1,16,power_law_1.01,0.19344768524169922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,1,16,power_law_1.01,0.44461822509765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,1,16,power_law_1.01,0.19451520919799806
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,1,16,power_law_1.01,3.900339050292969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,1,16,power_law_1.01,0.53731201171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,1,16,power_law_1.2,0.15568639755249022
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,1,16,power_law_1.01,0.8437490844726563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,1,16,power_law_1.2,0.32485633850097656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,1,16,power_law_1.01,0.9594252777099609
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,1,16,power_law_1.2,0.34015361785888676
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,1,16,power_law_1.2,0.3465779113769531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,1,16,power_law_1.01,0.19608959197998047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,1,16,power_law_1.2,0.28894208908081054
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,1,16,power_law_1.01,1.5457945251464844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,1,16,power_law_1.01,5.314962158203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,1,16,power_law_1.2,0.3335974502563476
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,1,16,power_law_1.2,0.35178752899169924
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,1,16,power_law_1.2,0.39634944915771486
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,1,16,power_law_1.2,0.4425830459594726
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,1,16,power_law_1.2,0.40602878570556644
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,1,16,power_law_1.2,0.4153510284423828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,1,16,power_law_1.2,0.3117657661437988
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,1,16,power_law_1.2,0.4373004913330078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,1,16,power_law_1.2,0.5024780654907227
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,1,16,power_law_1.2,0.34514816284179684
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,1,16,power_law_1.2,0.5846860885620118
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,1,16,power_law_1.2,0.6253401565551757
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,1,16,power_law_1.2,0.6621734619140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,1,16,power_law_1.01,2.0638131713867187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,1,16,power_law_1.2,0.8231398773193359
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,1,16,power_law_1.2,1.0493452453613281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,1,16,power_law_1.2,0.5256204986572266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,1,16,power_law_1.2,0.09601280212402344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,1,16,power_law_1.2,1.5633702087402344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,1,16,power_law_1.2,0.16564863204956054
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,1,16,power_law_1.2,0.15233280181884765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,1,16,power_law_1.2,1.712532501220703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,1,16,power_law_1.2,0.1636300849914551
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,1,16,power_law_1.2,0.1661350440979004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,1,16,power_law_1.2,0.14230655670166015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,1,16,power_law_1.2,0.15324543952941894
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,1,16,power_law_1.2,0.17201791763305666
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,1,16,power_law_1.2,0.17102848052978517
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,1,16,power_law_1.2,0.1777676773071289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,1,16,power_law_1.2,2.5201458740234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,1,16,power_law_1.2,0.18329343795776368
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,1,16,power_law_1.2,0.1842406463623047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,1,16,power_law_1.2,0.19251583099365235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,1,16,power_law_1.2,0.19402496337890623
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,1,16,power_law_1.2,0.1908121681213379
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,1,16,power_law_1.2,0.20807935714721681
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,1,16,power_law_1.2,0.2613529586791992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,1,16,power_law_1.2,0.2831027221679688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,1,16,power_law_1.2,3.3084686279296873
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,1,16,power_law_1.2,0.3344102478027344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,1,16,power_law_1.2,0.4073843383789063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,1,16,power_law_1.2,0.5501043319702148
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,1,16,power_law_1.2,0.19394304275512694
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,1,16,power_law_1.2,0.6683647918701172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,2,8,balanced,0.045099520683288576
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,1,16,power_law_1.2,4.3765567016601565
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,1,16,power_law_1.2,0.9741222381591796
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,2,8,balanced,0.058787841796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,2,8,balanced,0.07168896198272705
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,2,8,balanced,0.09664383888244629
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,2,8,balanced,0.15930368423461913
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,1,16,power_law_1.2,1.199741439819336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,2,8,balanced,0.35686912536621096
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,2,8,balanced,0.4527974319458008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,2,8,balanced,0.4600588989257813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,2,8,balanced,0.4638694381713867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,2,8,balanced,0.46441600799560545
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,1,16,power_law_1.2,1.7276197814941405
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,2,8,balanced,0.4614246368408203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,2,8,balanced,0.46914817810058596
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,2,8,balanced,0.469354248046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,2,8,balanced,0.4763673782348633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,1,16,power_law_1.2,2.2472154235839845
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,2,8,balanced,0.48543487548828124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,1,16,power_law_1.2,6.40445556640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,2,8,balanced,0.49550334930419926
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,2,8,balanced,0.511773452758789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,2,8,balanced,0.5472102355957031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,2,8,balanced,0.5871078491210937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,2,8,balanced,0.6558502197265625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,2,8,balanced,0.7177011108398437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,2,8,balanced,0.031481599807739256
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,2,8,balanced,0.0360806393623352
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,2,8,balanced,0.04565887928009033
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,2,8,balanced,1.027220458984375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,2,8,balanced,0.06310400009155273
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,2,8,balanced,0.10004480361938477
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,2,8,balanced,1.4229273986816406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,2,8,balanced,0.17413248062133788
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,2,8,balanced,0.1930009651184082
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,2,8,balanced,0.19536895751953126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,2,8,balanced,0.19723648071289063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,2,8,balanced,0.20064256668090819
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,2,8,balanced,1.801158447265625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,2,8,balanced,0.19142143249511717
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,2,8,balanced,0.8879103851318358
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,2,8,balanced,0.19442047119140624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,2,8,balanced,0.19631872177124024
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,2,8,balanced,0.2029862403869629
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,2,8,balanced,0.20826240539550783
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,2,8,balanced,0.21653760910034178
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,2,8,balanced,0.23059839248657227
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,2,8,balanced,0.2547980880737305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,2,8,balanced,0.2755955123901367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,2,8,balanced,0.31947391510009765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,2,8,balanced,0.35985790252685546
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,2,8,balanced,2.6540237426757813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,2,8,balanced,0.4538419342041015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,2,8,power_law_1.01,0.09459839820861816
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,2,8,power_law_1.01,0.15631232261657715
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,2,8,balanced,0.7811174774169922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,2,8,power_law_1.01,0.2001024055480957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,2,8,balanced,0.9611379241943359
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,2,8,balanced,3.380977783203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,2,8,power_law_1.01,0.24539007186889647
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,2,8,power_law_1.01,0.28558591842651365
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,2,8,power_law_1.01,0.31989503860473634
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,2,8,balanced,1.389803466796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,2,8,power_law_1.01,0.383581428527832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,2,8,power_law_1.01,0.41811584472656255
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,2,8,power_law_1.01,0.35960193634033205
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,2,8,balanced,0.5500339126586914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,2,8,power_law_1.01,0.3766783905029297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,2,8,power_law_1.01,0.4017241668701172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,2,8,balanced,1.8191795349121094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,2,8,power_law_1.01,0.4089100646972656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,2,8,power_law_1.01,0.4251302337646484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,2,8,power_law_1.01,0.4524748611450195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,2,8,power_law_1.01,0.43865726470947264
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,2,8,power_law_1.01,0.44340736389160157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,2,8,power_law_1.01,0.5287039947509766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,2,8,power_law_1.01,0.5904857635498046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,2,8,power_law_1.01,0.607322883605957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,2,8,power_law_1.01,0.6803033447265625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,2,8,power_law_1.01,0.7929535675048828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,2,8,power_law_1.01,0.061040639877319336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,2,8,power_law_1.01,1.0407193756103514
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,2,8,power_law_1.01,0.09610367774963378
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,2,8,power_law_1.01,1.226280975341797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,2,8,power_law_1.01,0.11381119728088378
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,2,8,power_law_1.01,0.13553152084350586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,2,8,power_law_1.01,0.14810751914978026
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,2,8,power_law_1.01,0.16608640670776367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,2,8,power_law_1.01,0.1764748764038086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,2,8,power_law_1.01,0.17132287979125976
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,2,8,power_law_1.01,1.8236569213867189
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,2,8,power_law_1.01,0.18007551193237306
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,2,8,power_law_1.01,0.18423551559448242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,2,8,power_law_1.01,0.18976512908935547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,2,8,power_law_1.01,0.17607807159423827
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,2,8,power_law_1.01,0.18268159866333006
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,2,8,power_law_1.01,0.18077056884765624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,2,8,power_law_1.01,2.4276800537109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,2,8,power_law_1.01,0.18642688751220704
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,2,8,power_law_1.01,0.18697216033935546
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,2,8,power_law_1.01,0.21259136199951173
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,2,8,power_law_1.01,0.22911231994628906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,2,8,power_law_1.01,0.2755148887634277
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,2,8,power_law_1.01,0.3177203178405762
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,2,8,power_law_1.01,0.35068416595458984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,2,8,power_law_1.01,0.46017921447753907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,2,8,power_law_1.01,0.5465203094482421
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,2,8,power_law_1.01,3.5476339721679686
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,2,8,power_law_1.01,0.75557373046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,2,8,power_law_1.2,0.15623552322387696
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,2,8,power_law_1.01,0.9569344329833985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,2,8,power_law_1.2,0.16233472824096679
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,2,8,power_law_1.2,0.19453567504882813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,2,8,power_law_1.01,1.3972837829589844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,2,8,power_law_1.2,0.28880384445190427
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,2,8,power_law_1.2,0.29572608947753903
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,2,8,power_law_1.2,0.09514240264892579
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,2,8,power_law_1.2,0.3942374420166016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,2,8,power_law_1.2,0.37914241790771486
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,2,8,power_law_1.2,0.40548095703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,2,8,power_law_1.2,0.4028889465332031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,2,8,power_law_1.2,0.40329345703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,2,8,power_law_1.2,0.4239347076416015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,2,8,power_law_1.2,0.45178367614746096
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,2,8,power_law_1.01,4.470509948730468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,2,8,power_law_1.2,0.48636928558349607
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,2,8,power_law_1.2,0.3959347152709961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,2,8,power_law_1.2,0.6094220733642578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,2,8,power_law_1.01,1.8571621704101564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,2,8,power_law_1.2,0.6422566223144531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,2,8,power_law_1.2,0.728958740234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,2,8,power_law_1.2,0.4149363327026367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,2,8,power_law_1.2,0.8690662384033203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,2,8,power_law_1.2,0.0616051197052002
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,2,8,power_law_1.2,1.1381798553466798
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,2,8,power_law_1.2,0.09428223609924316
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,2,8,power_law_1.2,0.10523520469665529
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,2,8,power_law_1.2,1.4518476867675782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,2,8,power_law_1.2,0.1262476825714111
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,2,8,power_law_1.2,0.15084799766540527
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,2,8,power_law_1.2,0.15000703811645508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,2,8,power_law_1.2,0.1676313591003418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,2,8,power_law_1.2,1.9158131408691408
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,2,8,power_law_1.2,0.5626419067382813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,2,8,power_law_1.2,0.17731327056884766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,2,8,power_law_1.2,0.18000511169433592
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,2,8,power_law_1.2,0.18138879776000977
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,2,8,power_law_1.2,0.18587007522583007
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,2,8,power_law_1.2,0.1809267234802246
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,2,8,power_law_1.2,0.1816281509399414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,2,8,power_law_1.2,0.1792844772338867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,2,8,power_law_1.2,0.19165056228637695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,2,8,power_law_1.2,0.21660160064697265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,2,8,power_law_1.2,0.2357913589477539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,2,8,power_law_1.2,0.28640768051147464
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,2,8,power_law_1.2,0.3422118377685547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,2,8,power_law_1.2,0.35795967102050785
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,2,8,power_law_1.2,0.4876697540283203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,2,8,power_law_1.2,0.18393600463867188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,2,8,power_law_1.2,0.5926911926269531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,2,8,power_law_1.2,2.8658712768554686
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,4,4,balanced,0.04294528007507324
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,4,4,balanced,0.056435198783874506
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,2,8,power_law_1.2,4.303999938964844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,2,8,power_law_1.2,0.8668211364746095
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,4,4,balanced,0.06844672203063965
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,4,4,balanced,0.09771648406982422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,4,4,balanced,0.15927552223205566
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,2,8,power_law_1.2,1.094013442993164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,4,4,balanced,0.35066368103027346
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,2,8,power_law_1.2,4.812019348144531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,4,4,balanced,0.35259902954101563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,4,4,balanced,0.4434944152832031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,4,4,balanced,0.44542591094970707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,4,4,balanced,0.44943359374999997
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,4,4,balanced,0.46118911743164065
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,2,8,power_law_1.2,1.7048883056640627
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,4,4,balanced,0.46667392730712887
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,4,4,balanced,0.47945983886718746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,4,4,balanced,0.4875660705566406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,4,4,balanced,0.5006771087646484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,4,4,balanced,0.3508863830566406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,2,8,power_law_1.2,2.115296020507812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,4,4,balanced,0.5315289688110352
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,4,4,balanced,0.5649779129028321
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,4,4,balanced,0.6340774536132813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,4,4,balanced,0.6988620758056641
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,4,4,balanced,0.4569472122192383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,4,4,balanced,0.02937727928161621
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,4,4,balanced,0.8814323425292969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,4,4,balanced,1.0283583831787109
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,4,4,balanced,0.04851712226867676
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,4,4,balanced,0.06526847839355468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,4,4,balanced,0.10516480445861816
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,4,4,balanced,1.4038656616210938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,4,4,balanced,0.17894912719726563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,4,4,balanced,0.18193279266357423
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,4,4,balanced,0.18450431823730468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,4,4,balanced,1.7931968688964843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,4,4,balanced,0.20473472595214845
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,4,4,balanced,0.03678591966629029
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,4,4,balanced,0.20518400192260744
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,4,4,balanced,0.2073638343811035
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,4,4,balanced,0.21038719177246096
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,4,4,balanced,0.2129062461853027
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,4,4,balanced,0.2187968063354492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,4,4,balanced,0.2224883270263672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,4,4,balanced,0.22840831756591795
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,4,4,balanced,0.24010623931884764
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,4,4,balanced,0.2624703979492188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,4,4,balanced,2.6049932861328124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,4,4,balanced,0.2812761688232422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,4,4,balanced,0.3234892654418945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,4,4,balanced,0.36399486541748044
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,4,4,balanced,0.44145790100097654
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,4,4,balanced,0.5273548889160156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,4,4,power_law_1.01,0.0668620777130127
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,4,4,balanced,3.337346496582031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,4,4,balanced,0.7736166381835938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,4,4,power_law_1.01,0.09401472091674805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,4,4,balanced,0.934150390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,4,4,power_law_1.01,0.11609472274780272
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,4,4,power_law_1.01,0.13603455543518067
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,4,4,power_law_1.01,0.18314367294311523
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,4,4,power_law_1.01,0.2121664047241211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,4,4,power_law_1.01,0.2978483200073242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,4,4,balanced,1.3543128967285156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,4,4,power_law_1.01,0.3104806327819824
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,4,4,power_law_1.01,0.3511372756958008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,4,4,power_law_1.01,0.3612966537475586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,4,4,power_law_1.01,0.3842009735107422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,4,4,power_law_1.01,0.36197120666503907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,4,4,power_law_1.01,0.4040383911132812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,4,4,balanced,1.781658935546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,4,4,power_law_1.01,0.4244569778442383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,4,4,power_law_1.01,0.41784320831298827
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,4,4,power_law_1.01,0.47517951965332034
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,4,4,power_law_1.01,0.5180902481079102
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,4,4,power_law_1.01,0.5711103820800781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,4,4,power_law_1.01,0.5893081665039063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,4,4,power_law_1.01,0.653276138305664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,4,4,power_law_1.01,0.7494258880615234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,4,4,power_law_1.01,0.047123198509216306
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,4,4,power_law_1.01,0.9708889770507814
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,4,4,power_law_1.01,0.06231808185577392
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,4,4,power_law_1.01,1.1878771209716796
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,4,4,power_law_1.01,0.07668992042541503
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,4,4,power_law_1.01,0.08979968070983887
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,4,4,power_law_1.01,0.12773119926452636
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,4,4,power_law_1.01,1.5914854431152343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,4,4,power_law_1.01,0.1412838363647461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,4,4,power_law_1.01,0.16331520080566406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,4,4,power_law_1.01,0.1817919921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,4,4,power_law_1.01,0.1964646339416504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,4,4,power_law_1.01,2.03928955078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,4,4,power_law_1.01,0.19983999252319334
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,4,4,power_law_1.01,0.1901811218261719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,4,4,power_law_1.01,0.19833599090576173
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,4,4,power_law_1.01,0.20248832702636718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,4,4,power_law_1.01,0.21354751586914061
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,4,4,power_law_1.01,0.22344703674316407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,4,4,power_law_1.01,0.15994239807128907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,4,4,power_law_1.01,0.2557619285583496
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,4,4,power_law_1.01,0.2769753646850586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,4,4,power_law_1.01,0.34811519622802733
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,4,4,power_law_1.01,0.18068607330322267
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,4,4,power_law_1.01,3.0317977905273437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,4,4,power_law_1.01,0.3874816131591797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,4,4,power_law_1.01,0.5084390258789062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,4,4,power_law_1.01,0.5955020904541015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,4,4,power_law_1.2,0.09277567863464356
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,4,4,power_law_1.01,0.8204019165039063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,4,4,power_law_1.2,0.10959872245788574
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,4,4,power_law_1.01,4.048458251953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,4,4,power_law_1.01,1.0377510070800782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,4,4,power_law_1.2,0.18515199661254883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,4,4,power_law_1.2,0.3156159973144531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,4,4,power_law_1.01,1.4811506652832032
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,4,4,power_law_1.2,0.0675878381729126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,4,4,power_law_1.2,0.3289190292358398
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,4,4,power_law_1.2,0.34153343200683595
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,4,4,power_law_1.2,0.3574220657348633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,4,4,power_law_1.01,2.030389709472656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,4,4,power_law_1.2,0.39500926971435546
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,4,4,power_law_1.2,0.11962623596191406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,4,4,power_law_1.2,0.17840511322021485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,4,4,power_law_1.2,0.43263233184814454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,4,4,power_law_1.2,0.43812606811523436
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,4,4,power_law_1.2,0.29318784713745116
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,4,4,power_law_1.2,0.4922995376586914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,4,4,power_law_1.2,0.549175033569336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,4,4,power_law_1.2,0.5836544036865234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,4,4,power_law_1.2,0.5919180679321289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,4,4,power_law_1.2,0.6927347564697266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,4,4,power_law_1.2,0.7849715423583985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,4,4,power_law_1.2,1.0127705383300782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,4,4,power_law_1.2,1.324418487548828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,4,4,power_law_1.2,0.047560958862304686
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,4,4,power_law_1.2,0.06278143882751465
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,4,4,power_law_1.2,0.4004953765869141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,4,4,power_law_1.2,0.07303167819976807
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,4,4,power_law_1.2,1.7490304565429688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,4,4,power_law_1.2,0.09222271919250488
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,4,4,power_law_1.2,0.11884672164916991
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,4,4,power_law_1.2,0.13558527946472168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,4,4,power_law_1.2,0.16301952362060548
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,4,4,power_law_1.2,0.162685432434082
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,4,4,power_law_1.2,0.18408063888549803
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,4,4,power_law_1.2,2.4188391113281247
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,4,4,power_law_1.2,0.19040128707885742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,4,4,power_law_1.2,0.1990540885925293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,4,4,power_law_1.2,0.18958335876464844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,4,4,power_law_1.2,0.20241535186767576
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,4,4,power_law_1.2,0.21911808013916018
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,4,4,power_law_1.2,0.234703369140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,4,4,power_law_1.2,0.26729856491088866
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,4,4,power_law_1.2,0.29983743667602536
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,4,4,power_law_1.2,3.246760864257813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,4,4,power_law_1.2,0.18299903869628906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,4,4,power_law_1.2,0.35543041229248046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,4,4,power_law_1.2,0.39469184875488283
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,4,4,power_law_1.2,0.5148352050781251
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,4,4,power_law_1.2,0.6272102355957031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,8,2,balanced,0.04728960037231446
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,8,2,balanced,0.056142082214355474
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,8,2,balanced,0.06836480140686035
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,4,4,power_law_1.2,4.290953063964844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,8,2,balanced,0.0961087989807129
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,8,2,balanced,0.16528127670288087
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,4,4,power_law_1.2,1.110909423828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,8,2,balanced,0.3517055892944336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,8,2,balanced,0.3519500732421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,8,2,balanced,0.3508198547363281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,8,2,balanced,0.35339134216308593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,4,4,power_law_1.2,1.6508799743652343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,8,2,balanced,0.3608665466308594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,4,4,power_law_1.2,0.8690060424804689
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,8,2,balanced,0.4440755081176758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,8,2,balanced,0.4502041625976562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,8,2,balanced,0.4653900909423828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,8,2,balanced,0.4584640121459961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,4,4,power_law_1.2,2.14071044921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,8,2,balanced,0.4727078247070312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,8,2,balanced,0.4841279983520508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,8,2,balanced,0.5171084976196288
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,4,4,power_law_1.2,0.20348928451538084
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,8,2,balanced,0.5440230560302735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,8,2,balanced,0.3569279861450195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,8,2,balanced,0.6076364898681641
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,8,2,balanced,0.671124496459961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,8,2,balanced,0.030495998859405515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,8,2,balanced,0.03878144025802612
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,8,2,balanced,0.049547519683837894
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,8,2,balanced,0.07119616031646728
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,8,2,balanced,1.073529586791992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,8,2,balanced,0.1178009605407715
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,8,2,balanced,0.2090790367126465
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,8,2,balanced,1.4378253173828124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,8,2,balanced,0.2103167915344238
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,8,2,balanced,0.21170944213867188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,8,2,balanced,0.21415552139282226
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,8,2,balanced,0.21671424865722658
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,8,2,balanced,1.8075379943847658
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,8,2,balanced,0.907240982055664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,8,2,balanced,0.25706880569458007
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,8,2,balanced,0.2592076873779297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,8,2,balanced,0.2633958435058594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,8,2,balanced,0.26728704452514646
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,8,2,balanced,0.2706841659545899
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,8,2,balanced,0.2774297523498535
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,8,2,balanced,0.2957209587097168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,8,2,balanced,0.3128767967224121
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,8,2,balanced,0.3484275054931641
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,8,2,balanced,2.6262603759765626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,8,2,balanced,0.38700927734375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,8,2,balanced,0.4620544052124023
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,8,2,balanced,0.5421670532226562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,8,2,power_law_1.01,0.05619071960449219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,8,2,balanced,3.3327398681640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,8,2,balanced,0.8276902770996093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,8,2,power_law_1.01,0.08179455757141113
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,8,2,balanced,0.21278207778930663
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,8,2,power_law_1.01,0.10948224067687988
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,8,2,power_law_1.01,0.15352191925048828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,8,2,power_law_1.01,0.17231744766235352
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,8,2,power_law_1.01,0.247064323425293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,8,2,balanced,1.4244979858398437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,8,2,power_law_1.01,0.2515788841247558
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,8,2,power_law_1.01,0.2669913673400879
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,8,2,power_law_1.01,0.27538944244384767
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,8,2,power_law_1.01,0.31558143615722656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,8,2,power_law_1.01,0.06836607933044433
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,8,2,power_law_1.01,0.3660518264770508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,8,2,balanced,1.8531097412109374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,8,2,power_law_1.01,0.37912960052490235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,8,2,balanced,0.9835648345947267
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,8,2,power_law_1.01,0.38421504974365234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,8,2,power_law_1.01,0.3926873779296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,8,2,power_law_1.01,0.42594558715820313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,8,2,power_law_1.01,0.4580953598022461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,8,2,power_law_1.01,0.5889766311645508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,8,2,power_law_1.01,0.5728704071044921
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,8,2,power_law_1.01,0.6418495941162109
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,8,2,power_law_1.01,0.7065087890624999
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,8,2,power_law_1.01,0.03714560031890869
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,8,2,power_law_1.01,0.912216339111328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,8,2,power_law_1.01,1.1108518218994141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,8,2,power_law_1.01,0.04872960090637207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,8,2,power_law_1.01,0.06092415809631348
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,8,2,power_law_1.01,0.08081664085388184
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,8,2,power_law_1.01,1.4709184265136719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,8,2,power_law_1.01,0.11329024314880372
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,8,2,power_law_1.01,0.1397862434387207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,8,2,power_law_1.01,0.16802944183349608
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,8,2,power_law_1.01,0.1752742385864258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,8,2,power_law_1.01,0.18135679244995118
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,8,2,power_law_1.01,1.9463987731933592
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,8,2,power_law_1.01,0.19108736038208007
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,8,2,power_law_1.01,0.22210048675537108
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,8,2,power_law_1.01,0.22401920318603516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,8,2,power_law_1.01,0.23334911346435544
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,8,2,power_law_1.01,0.2382476806640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,8,2,power_law_1.01,0.2466124725341797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,8,2,power_law_1.01,0.2611916732788086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,8,2,power_law_1.01,0.2857139205932617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,8,2,power_law_1.01,2.812140808105469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,8,2,power_law_1.01,0.31577215194702146
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,8,2,power_law_1.01,0.16868736267089843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,8,2,power_law_1.01,0.3810291290283203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,8,2,power_law_1.01,0.42895870208740233
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,8,2,power_law_1.01,0.5364518356323242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,8,2,power_law_1.2,0.05592063903808594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,8,2,power_law_1.01,0.653100814819336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,8,2,power_law_1.01,3.556529846191406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,8,2,power_law_1.2,0.06707839965820313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,8,2,power_law_1.2,0.07795199871063233
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,8,2,power_law_1.01,0.8688896179199219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,8,2,power_law_1.2,0.10442751884460448
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,8,2,power_law_1.2,0.1584550380706787
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,8,2,power_law_1.2,0.14163455963134766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,8,2,power_law_1.01,1.104167709350586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,8,2,power_law_1.2,0.23916160583496096
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,8,2,power_law_1.2,0.2585408020019531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,8,2,power_law_1.2,0.2538368034362793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,8,2,power_law_1.2,0.27870975494384764
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,8,2,power_law_1.2,0.3231935882568359
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,8,2,power_law_1.01,1.5558207702636717
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,8,2,power_law_1.2,0.3809203338623047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,8,2,power_law_1.2,0.3640230560302734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,8,2,power_law_1.2,0.39536510467529296
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,8,2,power_law_1.2,0.4140108871459961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,8,2,power_law_1.2,0.44013313293457035
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,8,2,power_law_1.2,0.47914112091064454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,8,2,power_law_1.01,2.044938201904297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,8,2,power_law_1.2,0.5851993560791016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,8,2,power_law_1.2,0.5952729415893555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,8,2,power_law_1.2,0.6548185729980469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,8,2,power_law_1.2,0.7449164581298828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,8,2,power_law_1.2,0.0364518404006958
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,8,2,power_law_1.2,0.9546521759033203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,8,2,power_law_1.2,0.04888192176818847
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,8,2,power_law_1.2,0.05785984039306641
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,8,2,power_law_1.2,1.158993911743164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,8,2,power_law_1.2,0.07482240200042725
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,8,2,power_law_1.2,0.10172287940979004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,8,2,power_law_1.2,0.12936063766479494
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,8,2,power_law_1.2,1.5050303649902343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,8,2,power_law_1.2,0.16435327529907226
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,8,2,power_law_1.2,0.17065471649169922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,8,2,power_law_1.2,0.17646591186523436
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,8,2,power_law_1.2,0.1766681671142578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,8,2,power_law_1.2,0.18432512283325195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,8,2,power_law_1.2,0.23069311141967774
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,8,2,power_law_1.2,2.0059584045410155
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,8,2,power_law_1.2,0.23601024627685546
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,8,2,power_law_1.2,0.23947391510009766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,8,2,power_law_1.2,0.2611123275756836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,8,2,power_law_1.2,0.2729395294189453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,8,2,power_law_1.2,0.29267967224121094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,8,2,power_law_1.2,0.34577919006347657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,8,2,power_law_1.2,0.3933900833129883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,8,2,power_law_1.2,0.44194561004638666
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,8,2,power_law_1.2,2.893592224121094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,8,2,power_law_1.2,0.565483512878418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,8,2,power_law_1.2,0.22492799758911133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,8,2,power_law_1.2,0.6777011108398437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,16,1,balanced,0.053902077674865725
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,16,1,balanced,0.0668012809753418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,16,1,balanced,0.07913152217864991
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,16,1,balanced,0.10839296340942384
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,8,2,power_law_1.2,0.8864588928222655
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,8,2,power_law_1.2,3.6674572753906247
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,16,1,balanced,0.17731327056884766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,8,2,power_law_1.2,1.1336678314208986
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,16,1,balanced,0.38251136779785155
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,16,1,balanced,0.38093822479248046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,16,1,balanced,0.38804481506347654
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,16,1,balanced,0.3891411209106445
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,8,2,power_law_1.2,1.61133056640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,16,1,balanced,0.38650047302246093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,16,1,balanced,0.392874870300293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,16,1,balanced,0.39954753875732424
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,16,1,balanced,0.4039590454101562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,16,1,balanced,0.41034622192382814
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,8,2,power_law_1.2,2.081347808837891
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,16,1,balanced,0.4863686370849609
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,16,1,balanced,0.4842438507080078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,16,1,balanced,0.49865791320800784
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,16,1,balanced,0.5247769546508789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,16,1,balanced,0.5537555313110352
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,16,1,balanced,0.6109056091308593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,16,1,balanced,0.6760076904296876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,16,1,balanced,0.04025599956512451
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,16,1,balanced,0.973193588256836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,16,1,balanced,1.0135884857177735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,16,1,balanced,0.04978240013122558
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,16,1,balanced,0.06539199829101563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,16,1,balanced,0.09357439994812011
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,16,1,balanced,0.1551961612701416
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,16,1,balanced,0.27978368759155275
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,16,1,balanced,1.391263427734375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,16,1,balanced,0.2819417572021484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,16,1,balanced,0.2854988861083984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,16,1,balanced,0.28711360931396485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,16,1,balanced,0.2900006484985352
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,16,1,balanced,1.7739442443847657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,16,1,balanced,0.2905388832092285
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,16,1,balanced,0.29521087646484373
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,16,1,balanced,0.29939008712768556
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,16,1,balanced,0.3649625778198242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,16,1,balanced,0.37292865753173826
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,16,1,balanced,0.28210624694824216
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,16,1,balanced,0.37173057556152345
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,16,1,balanced,0.38850879669189453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,16,1,balanced,0.40468673706054686
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,16,1,balanced,2.6154687500000002
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,16,1,balanced,0.4373766326904297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,16,1,balanced,0.4701862335205078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,16,1,balanced,0.5355129623413086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,16,1,balanced,0.6012684631347656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,16,1,power_law_1.01,0.04988671779632568
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,16,1,power_law_1.01,0.06632959842681885
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,16,1,power_law_1.01,0.07867968082427979
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,16,1,balanced,0.9564857482910156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,16,1,power_law_1.01,0.10864255905151368
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,16,1,balanced,3.4612939453125002
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,16,1,power_law_1.01,0.1463807964324951
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,16,1,balanced,1.0970355224609374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,16,1,power_law_1.01,0.1990969657897949
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,16,1,power_law_1.01,0.26012287139892576
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,16,1,power_law_1.01,0.27254144668579106
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,16,1,power_law_1.01,0.28301248550415037
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,16,1,power_law_1.01,0.29890815734863285
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,16,1,power_law_1.01,0.35476478576660153
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,16,1,balanced,1.616851806640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,16,1,power_law_1.01,0.375299186706543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,16,1,power_law_1.01,0.3841651153564453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,16,1,power_law_1.01,0.38904510498046874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,16,1,power_law_1.01,0.4218431854248047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,16,1,power_law_1.01,0.4344012832641601
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,16,1,balanced,2.1194476318359374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,16,1,power_law_1.01,0.47340225219726567
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,16,1,power_law_1.01,0.5703430557250977
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,16,1,power_law_1.01,0.6129516983032227
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,16,1,power_law_1.01,0.6547443389892578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,16,1,power_law_1.01,0.7181887817382813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,16,1,power_law_1.01,0.03982592105865478
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,16,1,power_law_1.01,0.9072844696044922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,16,1,power_law_1.01,0.048161921501159666
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,16,1,power_law_1.01,1.0837958526611327
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,16,1,power_law_1.01,0.06534656047821044
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,16,1,power_law_1.01,0.09425344467163085
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,16,1,power_law_1.01,0.12819840431213378
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,16,1,power_law_1.01,0.16622911453247072
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,16,1,power_law_1.01,0.2210905647277832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,16,1,power_law_1.01,1.4913369750976562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,16,1,power_law_1.01,0.2308742332458496
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,16,1,power_law_1.01,0.23363136291503905
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,16,1,power_law_1.01,0.2375168037414551
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,16,1,power_law_1.01,1.908211212158203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,16,1,power_law_1.01,0.24840768814086914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,16,1,power_law_1.01,0.25527679443359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,16,1,power_law_1.01,0.25949440002441404
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,16,1,power_law_1.01,0.26412351608276363
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,16,1,power_law_1.01,0.3293555068969726
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,16,1,power_law_1.01,0.3360627365112305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,16,1,power_law_1.01,0.3489273452758789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,16,1,power_law_1.01,0.37164161682128904
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,16,1,power_law_1.01,0.40631168365478515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,16,1,power_law_1.01,0.47449089050292964
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,16,1,power_law_1.01,2.719852294921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,16,1,power_law_1.01,0.5275884628295898
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,16,1,power_law_1.01,0.6344819259643555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,16,1,power_law_1.01,0.7714918518066407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,16,1,power_law_1.2,0.038492159843444826
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,16,1,power_law_1.01,1.0164230346679688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,16,1,power_law_1.2,0.06656256198883057
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,16,1,power_law_1.2,0.07927552223205567
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,16,1,power_law_1.2,0.10347455978393554
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,16,1,power_law_1.2,0.13949503898620605
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,16,1,power_law_1.01,1.269705581665039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,16,1,power_law_1.2,0.18948415756225584
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,16,1,power_law_1.2,0.2542515182495117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,16,1,power_law_1.01,3.5417553710937497
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,16,1,power_law_1.2,0.27506303787231445
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,16,1,power_law_1.01,1.7870361328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,16,1,power_law_1.2,0.30964031219482424
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,16,1,power_law_1.2,0.3585567855834961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,16,1,power_law_1.2,0.37745216369628903
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,16,1,power_law_1.2,0.38355838775634765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,16,1,power_law_1.2,0.39199550628662105
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,16,1,power_law_1.2,0.4389791870117188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,16,1,power_law_1.01,2.284393005371094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,16,1,power_law_1.2,0.2695840072631836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,16,1,power_law_1.2,0.4524095916748047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,16,1,power_law_1.2,0.489153938293457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,16,1,power_law_1.2,0.6044102478027343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,16,1,power_law_1.2,0.6229555130004882
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,16,1,power_law_1.2,0.6733805084228516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,16,1,power_law_1.2,0.7410578918457031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,16,1,power_law_1.2,0.03916224002838135
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,16,1,power_law_1.2,0.9307910156249999
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,16,1,power_law_1.2,0.048439040184020996
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,16,1,power_law_1.2,1.1210323333740235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,16,1,power_law_1.2,0.06617152214050293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,16,1,power_law_1.2,0.08894847869873047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,16,1,power_law_1.2,0.11985792160034178
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,16,1,power_law_1.2,1.5170040893554688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,16,1,power_law_1.2,0.15786368370056153
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,16,1,power_law_1.2,0.21717887878417969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,16,1,power_law_1.2,0.22636608123779295
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,16,1,power_law_1.2,0.23152959823608396
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,16,1,power_law_1.2,1.9265484619140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,16,1,power_law_1.2,0.23947519302368164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,16,1,power_law_1.2,0.24532863616943362
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,16,1,power_law_1.2,0.25233600616455076
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,16,1,power_law_1.2,0.2595417594909668
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,16,1,power_law_1.2,0.2693484878540039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,16,1,power_law_1.2,0.3360940933227539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,16,1,power_law_1.2,0.3463353729248047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,16,1,power_law_1.2,0.35471870422363283
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,16,1,power_law_1.2,0.39165889739990234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,16,1,power_law_1.2,0.43999870300292965
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,16,1,power_law_1.2,2.7557464599609376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,16,1,power_law_1.2,0.493480339050293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,16,1,power_law_1.2,0.5431615829467773
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,1,16,balanced,0.01844607949256897
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,16,1,power_law_1.2,0.7956057739257812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,1,16,balanced,0.018389760255813598
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,1,16,balanced,0.019173120260238645
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,1,16,balanced,0.03751424074172974
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,16,1,power_law_1.2,1.0265017700195311
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,16,1,power_law_1.2,3.578116455078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,1,16,balanced,0.05504127979278565
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,1,16,balanced,0.05807744026184082
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,1,16,balanced,0.06007296085357666
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,1,16,balanced,0.06002304077148437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,1,16,balanced,0.060993280410766605
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,1,16,balanced,0.06322944164276123
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,1,16,balanced,0.06572671890258788
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,1,16,balanced,0.06797823905944825
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,16,1,power_law_1.2,1.292211151123047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,1,16,balanced,0.05977600097656251
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,1,16,balanced,0.06230400085449219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,1,16,balanced,0.0626201581954956
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,1,16,balanced,0.0723737621307373
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,1,16,balanced,0.07737984180450438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,1,16,balanced,0.09323007583618163
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,1,16,balanced,0.026912000179290775
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,1,16,balanced,0.10147583961486817
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,1,16,balanced,0.13682815551757813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,1,16,balanced,0.16515327453613282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,1,16,balanced,0.2451263999938965
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,1,16,balanced,0.3009075164794922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,16,1,power_law_1.2,1.8079603576660155
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,1,16,balanced,0.021016321182250976
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,1,16,balanced,0.021094400882720948
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,1,16,balanced,0.4428556823730469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,1,16,balanced,0.021648640632629397
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,1,16,balanced,0.02183295965194702
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,1,16,balanced,0.024231679439544677
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,16,1,power_law_1.2,0.6612601470947266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,1,16,balanced,0.033286399841308593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,1,16,balanced,0.033712639808654785
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,1,16,balanced,0.034167039394378665
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,1,16,balanced,0.03478912115097046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,1,16,balanced,0.036771841049194336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,1,16,balanced,0.065763840675354
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,1,16,balanced,0.039141120910644533
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,1,16,balanced,0.04176511764526367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,16,1,power_law_1.2,2.314014739990234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,1,16,balanced,0.04155136108398437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,1,16,balanced,0.045459198951721194
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,1,16,balanced,0.05070720195770264
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,1,16,balanced,0.05611519813537598
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,1,16,balanced,0.06713215827941894
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,1,16,balanced,0.08025088310241699
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,1,16,balanced,0.10730239868164063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,1,16,balanced,0.13059840202331544
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,1,16,balanced,0.1893529510498047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,1,16,balanced,0.2336307144165039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,1,16,balanced,0.5692441558837891
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,1,16,balanced,0.3383603286743164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,1,16,balanced,0.03582335948944092
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,1,16,balanced,0.0382809591293335
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,1,16,power_law_1.01,0.036627199649810795
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,1,16,balanced,0.42995582580566405
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,1,16,power_law_1.01,0.036595199108123776
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,1,16,balanced,0.04278143882751465
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,1,16,power_law_1.01,0.0372326397895813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,1,16,power_law_1.01,0.055302400588989255
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,1,16,power_law_1.01,0.05890048027038575
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,1,16,power_law_1.01,0.056142082214355474
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,1,16,power_law_1.01,0.06279168128967286
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,1,16,power_law_1.01,0.06471168041229249
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,1,16,power_law_1.01,0.06586751937866211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,1,16,power_law_1.01,0.06409088134765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,1,16,power_law_1.01,0.06482816219329834
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,1,16,power_law_1.01,0.06498432159423828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,1,16,power_law_1.01,0.06747647762298584
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,1,16,power_law_1.01,0.07269375801086427
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,1,16,power_law_1.01,0.08741632461547852
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,1,16,power_law_1.01,0.10303744316101074
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,1,16,power_law_1.01,0.13509119987487792
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,1,16,power_law_1.01,0.1752934455871582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,1,16,power_law_1.01,0.20781440734863282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,1,16,power_law_1.01,0.2894540786743164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,1,16,power_law_1.01,0.047576317787170405
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,1,16,power_law_1.01,0.06139135837554932
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,1,16,power_law_1.01,0.39437950134277344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,1,16,power_law_1.01,0.02228991985321045
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,1,16,power_law_1.01,0.5861670303344726
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,1,16,power_law_1.01,0.023527679443359376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,1,16,power_law_1.01,0.02500607967376709
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,1,16,power_law_1.01,0.03261951923370361
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,1,16,power_law_1.01,0.03505408048629761
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,1,16,power_law_1.01,0.03758591890335083
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,1,16,power_law_1.01,0.03671040058135987
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,1,16,power_law_1.01,0.04173567771911621
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,1,16,power_law_1.01,0.04470143795013427
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,1,16,power_law_1.01,1.09531005859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,1,16,power_law_1.01,0.04601088047027588
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,1,16,power_law_1.01,0.04239232063293457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,1,16,power_law_1.01,0.044929280281066894
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,1,16,power_law_1.01,0.04626175880432129
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,1,16,power_law_1.01,0.04981247901916504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,1,16,power_law_1.01,0.02372607946395874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,1,16,power_law_1.01,0.0609600019454956
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,1,16,power_law_1.01,0.07135104179382325
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,1,16,power_law_1.01,0.09570816040039062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,1,16,power_law_1.01,0.1290329647064209
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,1,16,power_law_1.01,0.8570636749267578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,1,16,power_law_1.01,0.17419647216796874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,1,16,power_law_1.01,0.03676032066345215
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,1,16,power_law_1.01,0.21568511962890624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,1,16,power_law_1.01,0.3138201522827148
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,1,16,power_law_1.2,0.03636863946914673
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,1,16,power_law_1.2,0.035676159858703614
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,1,16,power_law_1.01,0.4029132843017578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,1,16,power_law_1.2,0.03666304111480713
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,1,16,power_law_1.2,0.03586303949356079
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,1,16,power_law_1.2,0.048360958099365234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,1,16,power_law_1.01,0.585687026977539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,1,16,power_law_1.2,0.05598720073699951
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,1,16,power_law_1.2,0.06017151832580566
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,1,16,power_law_1.01,0.7183411407470703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,1,16,power_law_1.2,0.06387199878692626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,1,16,power_law_1.2,0.06822656154632568
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,1,16,power_law_1.2,0.06841343879699707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,1,16,power_law_1.2,0.0628223991394043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,1,16,power_law_1.2,0.06324992179870606
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,1,16,power_law_1.2,0.06572159767150879
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,1,16,power_law_1.2,0.06878335952758789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,1,16,power_law_1.2,0.07532159805297851
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,1,16,power_law_1.2,0.09231616020202636
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,1,16,power_law_1.2,0.10966272354125976
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,1,16,power_law_1.2,0.1488319969177246
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,1,16,power_law_1.2,0.18234880447387697
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,1,16,power_law_1.2,0.28171903610229493
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,1,16,power_law_1.2,0.0577459192276001
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,1,16,power_law_1.2,0.06029695987701415
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,1,16,power_law_1.01,0.03336191892623901
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,1,16,power_law_1.2,0.022227199077606202
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,1,16,power_law_1.2,0.3779430389404297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,1,16,power_law_1.2,0.024526081085205077
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,1,16,power_law_1.2,0.023790080547332764
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,1,16,power_law_1.2,0.024075520038604737
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,1,16,power_law_1.2,0.03272063970565796
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,1,16,power_law_1.2,0.03409024000167847
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,1,16,power_law_1.2,0.034311680793762206
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,1,16,power_law_1.2,0.03500159978866577
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,1,16,power_law_1.2,0.03808511972427368
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,1,16,power_law_1.2,0.038314239978790285
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,1,16,power_law_1.01,0.03416960000991821
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,1,16,power_law_1.2,0.04333439826965332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,1,16,power_law_1.2,0.043632640838623046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,1,16,power_law_1.2,0.04212607860565186
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,1,16,power_law_1.2,0.04482304096221924
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,1,16,power_law_1.2,0.04925439834594726
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,1,16,power_law_1.2,0.05507584095001221
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,1,16,power_law_1.2,0.06266623973846436
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,1,16,power_law_1.2,0.07362559795379639
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,1,16,power_law_1.2,0.10412799835205078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,1,16,power_law_1.2,0.12973312377929686
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,1,16,power_law_1.2,0.16934143066406251
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,1,16,power_law_1.2,0.04413568019866944
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,1,16,power_law_1.2,0.7181183624267578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,1,16,power_law_1.2,0.2408025550842285
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,2,8,balanced,0.017511680126190185
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,1,16,power_law_1.2,0.3330342483520508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,2,8,balanced,0.017850879430770874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,2,8,balanced,0.018638080358505248
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,2,8,balanced,0.026035199165344237
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,2,8,balanced,0.03801471948623657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,1,16,power_law_1.2,0.5177267074584961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,2,8,balanced,0.05483520030975342
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,2,8,balanced,0.05996928215026855
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,2,8,balanced,0.06213119983673095
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,2,8,balanced,0.06594304084777833
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,2,8,balanced,0.05567488193511962
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,2,8,balanced,0.058393602371215825
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,2,8,balanced,0.059726080894470214
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,2,8,balanced,0.06206079959869385
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,2,8,balanced,0.06460544109344482
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,2,8,balanced,0.0659891176223755
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,2,8,balanced,0.07046656131744385
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,2,8,balanced,0.08204799652099609
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,2,8,balanced,0.09133952140808106
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,2,8,balanced,0.11634943962097169
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,2,8,balanced,0.05775743961334229
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,2,8,balanced,0.13401472091674804
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,2,8,balanced,0.184979190826416
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,2,8,balanced,0.23098880767822266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,2,8,balanced,0.33947647094726563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,2,8,balanced,0.42837760925292967
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,2,8,balanced,0.020625920295715333
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,2,8,balanced,0.6313151931762695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,2,8,balanced,0.02082175970077515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,2,8,balanced,0.0210752010345459
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,2,8,balanced,0.02189440011978149
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,2,8,balanced,0.8194866943359376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,2,8,balanced,0.0273638391494751
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,2,8,balanced,0.03895936012268066
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,2,8,balanced,0.04193408012390137
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,2,8,balanced,0.045758719444274905
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,2,8,balanced,0.04457856178283691
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,1,16,power_law_1.2,0.46993663787841794
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,2,8,balanced,0.03980799913406372
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,2,8,balanced,0.041277441978454586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,2,8,balanced,0.04349184036254883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,2,8,balanced,0.04508543968200683
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,2,8,balanced,0.0470579195022583
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,2,8,balanced,0.04862847805023193
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,2,8,balanced,0.05236991882324219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,2,8,balanced,0.06360960006713867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,2,8,balanced,0.07331711769104003
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,2,8,balanced,0.09597567558288575
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,2,8,balanced,0.11633024215698243
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,2,8,balanced,0.15964927673339843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,2,8,balanced,0.19647743225097655
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,2,8,balanced,0.03896320104598999
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,2,8,balanced,0.2915494346618652
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,2,8,balanced,0.3701375961303711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,2,8,power_law_1.01,0.022238719463348388
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,2,8,balanced,0.5412044906616211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,2,8,power_law_1.01,0.03129728078842163
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,2,8,power_law_1.01,0.030854399204254153
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,2,8,balanced,0.7009407806396484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,2,8,power_law_1.01,0.03586816072463989
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,2,8,power_law_1.01,0.03689728021621704
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,2,8,power_law_1.01,0.05165440082550049
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,2,8,power_law_1.01,0.05566080093383789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,2,8,power_law_1.01,0.05700607776641846
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,2,8,power_law_1.01,0.059171838760375975
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,2,8,power_law_1.01,0.06082176208496094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,2,8,power_law_1.01,0.06450560092926025
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,2,8,power_law_1.01,0.06125311851501465
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,2,8,power_law_1.01,0.06143616199493408
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,2,8,power_law_1.01,0.059619841575622556
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,2,8,power_law_1.01,0.06318719863891602
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,2,8,power_law_1.01,0.06529664039611817
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,2,8,power_law_1.01,0.06999040126800538
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,2,8,power_law_1.01,0.08036479949951172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,2,8,power_law_1.01,0.08910079956054687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,2,8,power_law_1.01,0.12370816230773926
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,2,8,power_law_1.01,0.14673791885375978
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,2,8,power_law_1.01,0.2018060874938965
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,2,8,power_law_1.01,0.2675289535522461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,2,8,power_law_1.01,0.3561548614501953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,2,8,power_law_1.01,0.5030643081665038
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,2,8,power_law_1.01,0.6686425781249999
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,2,8,power_law_1.01,0.021164801120758057
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,2,8,power_law_1.01,0.0248307204246521
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,2,8,power_law_1.01,0.9734272003173828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,2,8,power_law_1.01,0.026182401180267333
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,2,8,power_law_1.01,0.027165439128875736
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,2,8,power_law_1.01,0.03633152008056641
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,2,8,power_law_1.01,0.03859967947006225
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,2,8,power_law_1.01,0.04100351810455322
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,2,8,power_law_1.01,0.042917118072509766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,2,8,power_law_1.01,0.043438081741333005
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,2,8,power_law_1.01,0.04760191917419433
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,2,8,power_law_1.01,0.024858880043029784
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,2,8,power_law_1.01,0.04629759788513184
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,2,8,power_law_1.01,0.04302847862243652
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,2,8,power_law_1.01,0.04319488048553467
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,2,8,power_law_1.01,0.045519361495971675
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,2,8,power_law_1.01,0.04708223819732666
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,2,8,power_law_1.01,0.050321922302246094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,2,8,power_law_1.01,0.06298367977142334
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,2,8,power_law_1.01,0.07013504028320312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,2,8,power_law_1.01,0.09227904319763183
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,2,8,power_law_1.01,0.11368831634521484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,1,16,power_law_1.2,1.3396531677246093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,2,8,power_law_1.01,0.15993087768554687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,2,8,power_law_1.01,0.20852607727050781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,2,8,power_law_1.01,0.28420095443725585
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,2,8,power_law_1.2,0.020398080348968506
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,2,8,power_law_1.01,0.3730022430419922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,2,8,power_law_1.2,0.028528640270233153
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,2,8,power_law_1.2,0.0312550401687622
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,2,8,power_law_1.01,0.5150873565673828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,2,8,power_law_1.2,0.03597824096679687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,2,8,power_law_1.2,0.03570303916931152
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,2,8,power_law_1.01,0.6976831817626954
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,2,8,power_law_1.2,0.052596478462219244
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,2,8,power_law_1.2,0.05617023944854736
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,2,8,power_law_1.2,0.05708799839019776
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,2,8,power_law_1.2,0.05821055889129638
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,2,8,power_law_1.2,0.061365761756896973
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,2,8,power_law_1.2,0.06645376205444335
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,2,8,power_law_1.2,0.061109762191772464
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,2,8,power_law_1.2,0.05840000152587891
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,2,8,power_law_1.2,0.0615231990814209
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,2,8,power_law_1.2,0.0638643217086792
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,2,8,power_law_1.2,0.07147136211395264
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,2,8,power_law_1.2,0.08403455734252929
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,2,8,power_law_1.2,0.09965056419372559
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,2,8,power_law_1.2,0.13299967765808104
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,2,8,power_law_1.2,0.1596288013458252
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,2,8,power_law_1.2,0.22769664764404296
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,2,8,power_law_1.2,0.06709760189056396
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,1,16,power_law_1.2,1.010408935546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,2,8,power_law_1.2,0.28846719741821286
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,1,16,power_law_1.2,0.8872652435302735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,2,8,power_law_1.2,0.021269760131835937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,2,8,power_law_1.2,0.021542398929595946
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,2,8,power_law_1.2,0.02542207956314087
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,2,8,power_law_1.2,0.025090560913085935
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,2,8,power_law_1.2,0.026632959842681887
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,2,8,power_law_1.2,0.03859456062316895
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,2,8,power_law_1.2,0.03847935914993286
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,2,8,power_law_1.2,0.03964416027069092
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,2,8,power_law_1.2,0.04270847797393799
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,2,8,power_law_1.2,0.04813183784484863
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,2,8,power_law_1.2,0.04029952049255371
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,2,8,power_law_1.2,0.04138239860534668
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,2,8,power_law_1.2,0.045055999755859374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,2,8,power_law_1.2,0.0464140796661377
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,2,8,power_law_1.2,0.04919551849365235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,2,8,power_law_1.2,0.05167488098144531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,2,8,power_law_1.2,0.06538623809814453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,2,8,power_law_1.2,0.5501363372802734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,2,8,power_law_1.2,0.07347455978393555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,2,8,power_law_1.2,0.045068798065185545
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,2,8,power_law_1.2,0.09572735786437989
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,2,8,power_law_1.2,0.1188864040374756
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,2,8,power_law_1.2,0.170831356048584
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,2,8,power_law_1.2,0.42089214324951174
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,2,8,power_law_1.2,0.21056127548217773
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,4,4,balanced,0.018024959564208985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,4,4,balanced,0.018668800592422485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,4,4,balanced,0.020236799716949465
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,4,4,balanced,0.029418239593505858
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,4,4,balanced,0.04092927932739258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,4,4,balanced,0.043924479484558104
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,4,4,balanced,0.0565555191040039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,2,8,power_law_1.2,0.30998783111572265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,4,4,balanced,0.05930624008178711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,4,4,balanced,0.06215807914733886
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,4,4,balanced,0.06388351917266846
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,4,4,balanced,0.0556774377822876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,4,4,balanced,0.05701759815216064
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,4,4,balanced,0.05845248222351074
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,4,4,balanced,0.05999872207641601
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,4,4,balanced,0.06258560180664062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,4,4,balanced,0.06495488166809082
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,4,4,balanced,0.06869760036468506
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,4,4,balanced,0.07835519790649415
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,4,4,balanced,0.08825087547302246
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,4,4,balanced,0.11052032470703126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,4,4,balanced,0.1276364803314209
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,4,4,balanced,0.17398527145385742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,4,4,balanced,0.2217817687988281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,1,16,power_law_1.2,0.6840972900390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,4,4,balanced,0.3283366394042969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,4,4,balanced,0.41463550567626956
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,4,4,balanced,0.6125734329223633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,4,4,balanced,0.021239039897918702
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,4,4,balanced,0.8018240356445313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,4,4,balanced,0.02206592082977295
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,4,4,balanced,0.02392575979232788
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,4,4,balanced,0.030941441059112552
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,4,4,balanced,0.03375360012054444
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,4,4,balanced,0.04608767986297607
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,4,4,balanced,0.04776319980621338
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,4,4,balanced,0.049314560890197756
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,4,4,balanced,0.050807042121887205
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,4,4,balanced,0.04312191963195801
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,4,4,balanced,0.04538112163543701
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,4,4,balanced,0.04675712108612061
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,4,4,balanced,0.049190402030944824
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,2,8,power_law_1.2,0.40208000183105475
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,4,4,balanced,0.05222400188446045
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,4,4,balanced,0.05516416072845459
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,4,4,balanced,0.06522880077362062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,4,4,balanced,0.07463808059692383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,4,4,balanced,0.09372927665710448
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,4,4,balanced,0.10869376182556152
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,4,4,balanced,0.15660287857055663
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,4,4,balanced,0.1902310371398926
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,4,4,balanced,0.2799923133850098
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,4,4,balanced,0.05060991764068603
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,4,4,balanced,0.3580416107177734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,4,4,power_law_1.01,0.01901568055152893
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,4,4,balanced,0.5249638366699219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,2,8,power_law_1.2,0.5815296173095703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,4,4,power_law_1.01,0.02640000104904175
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,4,4,power_law_1.01,0.02904320001602173
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,4,4,power_law_1.01,0.031354880332946776
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,4,4,power_law_1.01,0.03550080060958862
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,4,4,power_law_1.01,0.040825600624084475
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,4,4,power_law_1.01,0.05641600131988526
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,4,4,power_law_1.01,0.05865471839904786
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,4,4,balanced,0.021704959869384765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,4,4,power_law_1.01,0.06355967998504639
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,4,4,power_law_1.01,0.06450560092926025
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,4,4,power_law_1.01,0.05654016017913819
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,4,4,power_law_1.01,0.05726079940795899
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,4,4,power_law_1.01,0.05982336044311524
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,4,4,power_law_1.01,0.06360832214355469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,4,4,power_law_1.01,0.06707200050354004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,4,4,power_law_1.01,0.07149951934814454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,2,8,power_law_1.2,1.0075917053222656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,4,4,balanced,0.6936780548095702
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,4,4,power_law_1.01,0.07950208187103272
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,4,4,power_law_1.01,0.08956159591674805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,4,4,power_law_1.01,0.11350272178649903
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,4,4,power_law_1.01,0.13910271644592284
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,4,4,power_law_1.01,0.19126144409179688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,4,4,power_law_1.01,0.06064511775970459
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,4,4,power_law_1.01,0.2588863945007324
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,4,4,power_law_1.01,0.3636812973022461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,4,4,power_law_1.01,0.02330496072769165
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,4,4,power_law_1.01,0.47106304168701174
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,4,4,power_law_1.01,0.024953598976135256
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,4,4,power_law_1.01,0.6390041732788087
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,4,4,power_law_1.01,0.027569921016693116
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,4,4,power_law_1.01,0.033354239463806154
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,4,4,power_law_1.01,0.04669055938720703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,4,4,power_law_1.01,0.049141759872436526
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,2,8,power_law_1.2,0.8394879913330078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,4,4,power_law_1.01,0.04909183979034424
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,4,4,power_law_1.01,0.04387072086334228
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,4,4,power_law_1.01,0.045829119682312014
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,4,4,power_law_1.01,0.047571201324462895
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,4,4,power_law_1.01,0.05056511878967285
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,4,4,power_law_1.01,0.05217152118682862
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,4,4,power_law_1.01,0.05632512092590332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,4,4,power_law_1.01,0.8409548950195311
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,4,4,power_law_1.01,0.06690815925598145
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,4,4,power_law_1.01,0.07609856128692627
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,4,4,power_law_1.01,0.1243238353729248
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,4,4,power_law_1.01,0.169303035736084
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,4,4,power_law_1.01,0.21325183868408204
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,4,4,power_law_1.01,0.3040435218811035
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,4,4,power_law_1.01,0.39984127044677736
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,4,4,power_law_1.01,0.5803302383422851
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,4,4,power_law_1.2,0.01891327977180481
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,2,8,power_law_1.2,0.754103012084961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,4,4,power_law_1.01,0.021629440784454345
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,4,4,power_law_1.2,0.02448512077331543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,4,4,power_law_1.01,0.7614579010009765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,4,4,power_law_1.2,0.028709120750427246
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,4,4,power_law_1.2,0.031306240558624264
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,4,4,power_law_1.2,0.03472383975982666
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,4,4,power_law_1.01,0.030396161079406736
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,4,4,power_law_1.2,0.040825600624084475
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,4,4,power_law_1.2,0.059311361312866205
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,4,4,power_law_1.2,0.059795198440551756
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,4,4,power_law_1.2,0.062152957916259764
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,4,4,power_law_1.2,0.06339456081390381
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,4,4,power_law_1.2,0.056604161262512206
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,4,4,power_law_1.2,0.05841919898986816
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,4,4,power_law_1.2,0.061644802093505854
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,4,4,power_law_1.2,0.06439680099487305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,4,4,power_law_1.2,0.06757760047912598
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,4,4,power_law_1.2,0.07177855968475341
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,4,4,power_law_1.2,0.08162176132202148
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,4,4,power_law_1.2,0.09190400123596192
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,4,4,power_law_1.2,0.11719679832458496
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,4,4,power_law_1.2,0.14302080154418945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,4,4,power_law_1.2,0.05633664131164551
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,4,4,power_law_1.2,0.2063884735107422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,4,4,power_law_1.2,0.26197631835937496
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,4,4,power_law_1.01,0.043854079246520995
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,4,4,power_law_1.01,0.04916736125946045
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,4,4,power_law_1.2,0.02173439979553223
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,4,4,power_law_1.2,0.023180799484252928
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,4,4,power_law_1.2,0.02438271999359131
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,4,4,power_law_1.2,0.02753151893615723
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,4,4,power_law_1.2,0.02907392024993897
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,4,4,power_law_1.2,0.03431551933288574
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,4,4,power_law_1.2,0.044026880264282225
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,4,4,power_law_1.2,0.04626304149627686
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,4,4,power_law_1.2,0.04880512237548828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,4,4,power_law_1.2,0.050044159889221194
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,4,4,power_law_1.2,0.04517632007598877
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,4,4,power_law_1.01,0.09751935958862304
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,4,4,power_law_1.2,0.04632832050323486
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,4,4,power_law_1.2,0.04931072235107422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,4,4,power_law_1.2,0.05209983825683594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,4,4,power_law_1.2,0.053738241195678715
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,4,4,power_law_1.2,0.057100801467895514
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,4,4,power_law_1.2,0.06822783946990967
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,4,4,power_law_1.2,0.07738111972808838
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,4,4,power_law_1.2,0.1011302375793457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,4,4,power_law_1.2,0.13137663841247557
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,4,4,power_law_1.2,0.17268096923828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,4,4,power_law_1.2,0.04483327865600586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,4,4,power_law_1.2,0.4786816024780274
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,4,4,power_law_1.2,0.21603839874267577
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,1,16,balanced,0.03883775949478149
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,1,16,balanced,0.04020736217498779
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,1,16,balanced,0.049628162384033205
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,4,4,power_law_1.2,0.380555534362793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,1,16,balanced,0.06030848026275635
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,1,16,balanced,0.08732159614562988
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,1,16,balanced,0.12360320091247559
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,1,16,balanced,0.1231167984008789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,1,16,balanced,0.12741248130798338
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,1,16,balanced,0.12799743652343748
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,1,16,balanced,0.13134464263916015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,1,16,balanced,0.13199872016906739
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,1,16,balanced,0.13711615562438967
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,1,16,balanced,0.14244864463806153
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,1,16,balanced,0.13496959686279297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,1,16,balanced,0.1381926441192627
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,1,16,balanced,0.14289279937744143
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,1,16,balanced,0.14995712280273438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,1,16,balanced,0.1638912010192871
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,4,4,power_law_1.2,0.3080729675292969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,1,16,balanced,0.17206783294677735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,1,16,balanced,0.23102592468261718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,1,16,balanced,0.26823040008544924
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,1,16,balanced,0.36392574310302733
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,1,16,balanced,0.4445951843261719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,4,4,power_law_1.2,0.40431488037109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,1,16,balanced,0.6425062561035156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,1,16,balanced,0.027132160663604736
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,1,16,balanced,0.8081497955322267
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,1,16,balanced,0.027361280918121338
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,1,16,balanced,0.0308787202835083
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,1,16,balanced,0.0407206392288208
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,1,16,balanced,0.05827072143554688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,1,16,balanced,0.06631423950195312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,1,16,balanced,1.1986854553222657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,1,16,balanced,0.06577280044555664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,1,16,balanced,0.07022079944610596
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,1,16,balanced,0.07130368232727051
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,1,16,balanced,0.07092480182647705
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,1,16,balanced,1.5612684631347657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,1,16,balanced,0.07412864208221435
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,1,16,balanced,0.07936384201049804
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,1,16,balanced,0.08357888221740722
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,1,16,balanced,0.07637119770050049
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,1,16,balanced,0.08304767608642578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,1,16,balanced,0.08849151611328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,1,16,balanced,0.09749631881713867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,1,16,balanced,0.1054207992553711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,1,16,balanced,0.1299174404144287
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,1,16,balanced,0.15128959655761717
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,1,16,balanced,0.20655231475830077
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,1,16,balanced,0.2448358345031738
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,1,16,balanced,0.3487948989868164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,1,16,balanced,0.42954238891601565
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,1,16,balanced,0.07900032043457031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,1,16,balanced,0.6343603134155273
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,1,16,power_law_1.01,0.08638079643249512
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,1,16,power_law_1.01,0.08377216339111329
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,1,16,power_law_1.01,0.08271360397338867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,1,16,power_law_1.01,0.0835814380645752
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,1,16,power_law_1.01,0.10478207588195801
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,1,16,power_law_1.01,0.11892607688903808
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,1,16,power_law_1.01,0.12164095878601074
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,1,16,power_law_1.01,0.12318592071533203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,1,16,power_law_1.01,0.12762880325317383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,1,16,power_law_1.01,0.12812031745910643
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,1,16,power_law_1.01,0.13450752258300783
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,1,16,power_law_1.01,0.13957759857177734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,1,16,balanced,0.8210316467285157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,1,16,power_law_1.01,0.14005248069763182
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,1,16,power_law_1.01,0.14606080055236817
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,1,16,power_law_1.01,0.15427583694458008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,1,16,power_law_1.01,0.16248064041137694
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,1,16,power_law_1.01,0.18339967727661133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,1,16,power_law_1.01,0.23747583389282229
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,1,16,power_law_1.01,0.2859942436218262
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,1,16,power_law_1.01,0.37361793518066405
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,1,16,power_law_1.01,0.46777599334716796
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,1,16,power_law_1.01,0.6359308624267579
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,1,16,power_law_1.01,0.8383424377441406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,1,16,power_law_1.01,1.233743362426758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,1,16,power_law_1.01,0.05655935764312744
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,1,16,power_law_1.01,1.6450457763671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,1,16,power_law_1.01,0.05560448169708252
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,1,16,power_law_1.01,0.057427201271057125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,1,16,power_law_1.01,0.046995201110839845
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,1,16,power_law_1.01,0.05557631969451904
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,1,16,power_law_1.01,0.06012800216674805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,4,4,power_law_1.2,0.7549529266357422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,1,16,power_law_1.01,2.4307391357421873
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,1,16,power_law_1.01,0.06451456069946289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,1,16,power_law_1.01,0.07022208213806153
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,1,16,power_law_1.01,0.07371263980865479
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,1,16,power_law_1.01,0.07816832065582276
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,1,16,power_law_1.01,0.07558271884918213
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,1,16,power_law_1.01,0.07954944133758544
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,1,16,power_law_1.01,0.08446847915649414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,1,16,power_law_1.01,0.09004287719726563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,1,16,power_law_1.01,0.09855487823486328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,1,16,power_law_1.01,0.0636672019958496
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,1,16,power_law_1.01,0.1311398410797119
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,1,16,power_law_1.01,0.1521945571899414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,1,16,power_law_1.01,0.07183743953704834
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,1,16,power_law_1.01,0.18642047882080076
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,1,16,power_law_1.01,0.2272038459777832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,1,16,power_law_1.01,0.31888256072998045
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,1,16,power_law_1.01,0.40770687103271486
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,1,16,power_law_1.01,0.5960934448242188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,4,4,power_law_1.2,0.9489202880859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,1,16,power_law_1.01,0.7696870422363281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,1,16,power_law_1.2,0.08650752067565917
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,1,16,power_law_1.01,3.6995559692382813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,1,16,power_law_1.2,0.08295295715332032
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,1,16,power_law_1.2,0.06596479892730714
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,1,16,power_law_1.01,1.0948006439208986
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,1,16,power_law_1.2,0.11495039939880372
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,1,16,power_law_1.2,0.12156543731689454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,1,16,power_law_1.2,0.12126720428466797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,1,16,power_law_1.2,0.128852481842041
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,1,16,power_law_1.2,0.13329919815063476
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,1,16,power_law_1.2,0.14006784439086914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,1,16,power_law_1.2,0.1404697608947754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,1,16,power_law_1.01,1.6730227661132813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,1,16,power_law_1.2,0.14247424125671387
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,1,16,power_law_1.2,0.07849088191986084
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,1,16,power_law_1.2,0.1475430393218994
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,1,16,power_law_1.2,0.1625587272644043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,1,16,power_law_1.2,0.2047078323364258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,1,16,power_law_1.2,0.2584076881408691
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,1,16,power_law_1.2,0.11388928413391113
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,1,16,power_law_1.2,0.31066879272460934
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,1,16,power_law_1.2,0.4461337661743164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,1,16,power_law_1.2,0.5555865478515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,4,4,power_law_1.2,0.7784627532958984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,1,16,power_law_1.2,0.77272705078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,1,16,power_law_1.2,0.17283199310302735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,1,16,power_law_1.2,0.05454720020294189
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,1,16,power_law_1.2,0.05388160228729248
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,1,16,power_law_1.2,0.05474175930023193
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,1,16,power_law_1.2,1.0745970916748047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,1,16,power_law_1.2,0.04578432083129883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,1,16,power_law_1.2,0.05257984161376953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,1,16,power_law_1.2,1.6521881103515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,1,16,power_law_1.2,0.061582078933715825
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,1,16,power_law_1.2,0.06209792137145996
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,1,16,power_law_1.2,0.06448128223419189
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,1,16,power_law_1.2,0.07224832057952882
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,1,16,power_law_1.2,0.07398784160614014
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,1,16,power_law_1.2,0.08092032432556152
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,1,16,power_law_1.2,0.07739776134490967
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,1,16,power_law_1.2,0.08394623756408691
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,1,16,power_law_1.2,0.0884607982635498
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,1,16,power_law_1.2,0.09409919738769532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,1,16,power_law_1.2,0.10085247993469237
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,1,16,power_law_1.2,0.1396787166595459
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,1,16,power_law_1.2,0.16046207427978515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,4,4,power_law_1.2,0.5862527847290039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,1,16,power_law_1.2,0.2762841606140137
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,1,16,power_law_1.2,0.3637503814697266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,1,16,power_law_1.2,0.07803775787353515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,1,16,power_law_1.2,0.19963775634765626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,1,16,power_law_1.2,0.4764723205566407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,2,8,balanced,0.024311039447784424
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,2,8,balanced,0.038567678928375246
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,1,16,power_law_1.2,0.6333670425415039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,2,8,balanced,0.046869759559631345
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,2,8,balanced,0.059202561378479
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,2,8,balanced,0.0933619213104248
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,2,8,balanced,0.12514176368713378
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,2,8,balanced,0.13092608451843263
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,2,8,balanced,0.1343654441833496
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,2,8,balanced,0.1362880039215088
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,2,8,balanced,0.14020735740661622
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,2,8,balanced,0.13110015869140626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,2,8,balanced,0.13357439994812012
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,2,8,balanced,0.1367039966583252
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,2,8,balanced,0.14324095726013183
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,2,8,balanced,0.14836352348327636
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,2,8,balanced,0.15361023902893067
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,2,8,balanced,0.16409984588623047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,2,8,balanced,0.1846963119506836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,2,8,balanced,0.19850751876831055
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,2,8,balanced,0.268719367980957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,2,8,balanced,0.31911935806274416
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,2,8,balanced,0.437336311340332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,1,16,power_law_1.2,2.0041317749023437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,2,8,balanced,0.5393228912353516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,1,16,power_law_1.2,0.9902681732177735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,2,8,balanced,0.7868863677978515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,1,16,power_law_1.2,2.1294822692871094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,2,8,balanced,0.025552639961242674
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,2,8,balanced,0.02630784034729004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,2,8,balanced,0.02965759992599487
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,2,8,balanced,0.9967040252685546
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,2,8,balanced,0.04191872119903565
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,2,8,balanced,0.061743359565734866
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,1,16,power_law_1.2,1.4375552368164062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,2,8,balanced,0.06982912063598633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,2,8,balanced,0.07252607822418214
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,2,8,balanced,0.0757209587097168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,2,8,balanced,1.4856588745117187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,2,8,balanced,0.07819519996643066
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,2,8,balanced,0.074170880317688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,2,8,balanced,0.07701632022857666
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,2,8,balanced,0.08292096138000489
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,2,8,balanced,0.08611328124999999
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,2,8,balanced,0.09012864112854005
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,2,8,balanced,0.10027392387390137
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,2,8,balanced,0.11850367546081544
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,2,8,balanced,0.13604607582092285
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,2,8,balanced,0.16837120056152344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,2,8,balanced,0.2010163116455078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,2,8,balanced,0.28594432830810546
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,2,8,balanced,0.3501887893676758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,2,8,balanced,0.08005760192871095
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,2,8,balanced,0.5076710510253906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,2,8,balanced,0.6419225311279296
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,2,8,power_law_1.01,0.05761407852172852
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,2,8,power_law_1.01,0.06504703998565674
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,2,8,power_law_1.01,0.07321983814239502
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,2,8,balanced,1.9470860290527345
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,2,8,balanced,0.954351348876953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,2,8,power_law_1.01,0.0704307222366333
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,2,8,power_law_1.01,0.06974720001220704
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,2,8,power_law_1.01,0.11249279975891113
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,2,8,power_law_1.01,0.12333696365356446
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,2,8,power_law_1.01,0.12911104202270507
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,2,8,balanced,0.07928832054138184
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,2,8,power_law_1.01,0.13063167572021483
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,2,8,power_law_1.01,0.13745408058166503
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,2,8,power_law_1.01,0.13351552009582518
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,2,8,power_law_1.01,0.13155200004577636
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,2,8,power_law_1.01,0.14502016067504883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,2,8,power_law_1.01,0.1537996768951416
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,2,8,power_law_1.01,0.15912832260131834
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,2,8,power_law_1.01,0.16990848541259768
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,2,8,power_law_1.01,0.20749183654785158
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,2,8,power_law_1.01,0.2477657508850098
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,2,8,power_law_1.01,0.35993343353271484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,2,8,balanced,1.241674270629883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,2,8,power_law_1.01,0.4119385528564453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,2,8,power_law_1.01,0.5403507232666016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,2,8,power_law_1.01,0.7173964691162109
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,2,8,power_law_1.01,0.040298237800598144
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,2,8,power_law_1.01,1.0977471923828124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,2,8,power_law_1.01,0.046012158393859866
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,2,8,power_law_1.01,0.11700096130371093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,2,8,power_law_1.01,1.2524658966064455
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,2,8,power_law_1.01,0.050677762031555176
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,2,8,power_law_1.01,0.049616641998291015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,2,8,power_law_1.01,0.04984576225280762
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,2,8,power_law_1.01,0.0660864019393921
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,2,8,power_law_1.01,0.06906367778778076
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,2,8,power_law_1.01,0.07088640213012695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,2,8,power_law_1.01,0.0717964792251587
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,2,8,power_law_1.01,0.07644415855407714
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,2,8,power_law_1.01,0.07358335971832275
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,2,8,power_law_1.01,0.07438464164733886
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,2,8,power_law_1.01,2.168145904541016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,2,8,power_law_1.01,0.08182656288146972
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,2,8,power_law_1.01,0.08867072105407715
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,2,8,power_law_1.01,0.09523839950561523
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,2,8,power_law_1.01,0.12195327758789062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,2,8,power_law_1.01,0.06336895942687988
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,2,8,power_law_1.01,2.8224359130859376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,2,8,power_law_1.01,0.14233471870422362
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,2,8,power_law_1.01,0.18197504043579102
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,2,8,power_law_1.01,0.22037248611450194
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,2,8,power_law_1.01,0.07815807819366455
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,2,8,power_law_1.01,0.292093448638916
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,2,8,power_law_1.01,0.3559334564208984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,2,8,power_law_1.01,0.5239116668701171
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,2,8,power_law_1.2,0.057968640327453615
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,2,8,power_law_1.01,0.6607564544677734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,2,8,power_law_1.2,0.054585599899291994
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,2,8,power_law_1.2,0.06991615772247314
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,2,8,power_law_1.2,0.07227392196655273
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,2,8,power_law_1.2,0.0662553596496582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,2,8,power_law_1.2,0.10889216423034669
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,2,8,power_law_1.01,1.047330551147461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,2,8,power_law_1.2,0.11855104446411133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,2,8,power_law_1.2,0.12407808303833008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,2,8,power_law_1.2,0.13125247955322267
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,2,8,power_law_1.2,0.13223808288574218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,2,8,power_law_1.2,0.1386828804016113
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,1,16,power_law_1.2,3.4588134765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,2,8,power_law_1.2,0.1440627193450928
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,2,8,power_law_1.2,0.1440460777282715
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,2,8,power_law_1.2,0.14755711555480958
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,2,8,power_law_1.2,0.1501363182067871
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,2,8,power_law_1.2,0.18247552871704104
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,2,8,power_law_1.2,0.26977279663085935
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,2,8,power_law_1.2,0.3740006256103515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,2,8,power_law_1.2,0.43671550750732424
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,2,8,power_law_1.2,0.6328665542602538
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,2,8,power_law_1.01,1.4403353881835936
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,2,8,power_law_1.2,0.15776639938354492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,2,8,power_law_1.2,0.8401497650146486
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,2,8,power_law_1.2,0.21495040893554687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,2,8,power_law_1.2,0.03938175916671753
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,2,8,power_law_1.2,0.042698240280151366
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,2,8,power_law_1.2,0.049236478805541994
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,2,8,power_law_1.2,1.243345947265625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,2,8,power_law_1.2,0.04962048053741455
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,2,8,power_law_1.2,0.04423423767089844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,2,8,power_law_1.2,0.0636953592300415
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,2,8,power_law_1.2,0.06655104160308838
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,2,8,power_law_1.2,0.06677760124206543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,2,8,power_law_1.2,0.07222015857696533
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,2,8,power_law_1.2,0.07532671928405762
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,2,8,power_law_1.2,0.07921023845672608
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,2,8,power_law_1.2,0.07766143798828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,2,8,power_law_1.2,0.07623167991638183
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,2,8,power_law_1.2,0.08034303665161133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,2,8,power_law_1.2,0.08383359909057617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,2,8,power_law_1.2,0.08765952110290527
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,2,8,power_law_1.2,0.09634816169738769
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,2,8,power_law_1.2,0.1290726375579834
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,2,8,power_law_1.2,0.15301247596740725
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,2,8,power_law_1.2,0.18888063430786134
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,2,8,power_law_1.2,0.22959360122680664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,2,8,power_law_1.2,0.3409062576293945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,2,8,power_law_1.2,1.7796044921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,2,8,power_law_1.2,0.41028865814208987
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,4,4,balanced,0.024270079135894775
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,4,4,balanced,0.03750400066375732
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,2,8,power_law_1.2,2.4300965881347656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,4,4,balanced,0.04641280174255371
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,4,4,balanced,0.061656317710876464
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,4,4,balanced,0.09148544311523438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,4,4,balanced,0.09597951889038085
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,4,4,balanced,0.12811903953552245
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,4,4,balanced,0.13126912117004394
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,4,4,balanced,0.13617152214050293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,4,4,balanced,0.12755583763122558
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,4,4,balanced,0.129367036819458
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,4,4,balanced,0.13208831787109374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,4,4,balanced,0.13586432456970216
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,4,4,balanced,0.13390975952148437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,4,4,balanced,0.14004096031188965
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,4,4,balanced,0.14501248359680177
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,4,4,balanced,0.15243519783020018
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,4,4,balanced,0.1709440040588379
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,4,4,balanced,0.18497407913208008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,4,4,balanced,0.25573888778686527
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,2,8,power_law_1.2,0.6000384140014648
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,4,4,balanced,0.2970534324645996
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,4,4,balanced,0.42215423583984374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,4,4,balanced,0.528664321899414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,4,4,balanced,0.7628160095214843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,4,4,balanced,0.9675583648681642
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,2,8,power_law_1.2,0.7782246398925781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,4,4,balanced,0.025542399883270263
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,4,4,balanced,0.026155519485473632
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,4,4,balanced,0.032016639709472654
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,4,4,balanced,0.044592638015747074
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,2,8,power_law_1.2,1.1737881469726563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,4,4,balanced,0.06554880142211914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,4,4,balanced,0.06793344020843506
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,4,4,balanced,0.07890560150146483
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,4,4,balanced,0.08131711959838868
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,4,4,balanced,0.08401535987854004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,4,4,balanced,0.08597248077392579
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,4,4,balanced,0.07789311885833741
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,4,4,balanced,0.08033280372619629
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,4,4,balanced,1.8809432983398438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,4,4,balanced,0.0819264030456543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,4,4,balanced,0.08516480445861815
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,4,4,balanced,0.08824447631835938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,4,4,balanced,0.09224703788757324
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,4,4,balanced,0.10116095542907715
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,4,4,balanced,1.4339109802246095
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,4,4,balanced,0.11762175559997559
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,4,4,balanced,0.13128959655761718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,4,4,balanced,0.16120576858520508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,4,4,balanced,0.1911257553100586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,4,4,balanced,0.28027135848999024
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,4,4,balanced,0.3382156753540039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,4,4,balanced,0.4932108688354492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,4,4,balanced,0.62076416015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,4,4,power_law_1.01,0.04611968040466309
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,4,4,power_law_1.01,0.049491200447082516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,4,4,power_law_1.01,0.05696000099182129
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,4,4,balanced,0.9254873657226563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,4,4,power_law_1.01,0.06321536064147949
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,4,4,power_law_1.01,0.06733952045440675
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,4,4,power_law_1.01,0.08194175720214844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,2,8,power_law_1.2,3.4397515869140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,4,4,power_law_1.01,0.11682815551757812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,4,4,balanced,1.201945571899414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,4,4,power_law_1.01,0.12196096420288086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,4,4,power_law_1.01,0.12523903846740722
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,4,4,power_law_1.01,0.12829695701599123
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,4,4,power_law_1.01,0.13237759590148926
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,4,4,power_law_1.01,0.1356608009338379
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,4,4,power_law_1.01,0.14126720428466796
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,4,4,power_law_1.01,0.14578816413879395
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,4,4,power_law_1.01,0.15532416343688965
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,4,4,power_law_1.01,0.1798784065246582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,4,4,power_law_1.01,0.21579904556274415
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,4,4,power_law_1.01,0.2900799942016602
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,4,4,power_law_1.01,0.37664894104003904
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,4,4,power_law_1.01,0.5097049713134766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,4,4,power_law_1.01,0.1270361614227295
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,4,4,power_law_1.01,0.6346611022949219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,4,4,power_law_1.01,0.1269337558746338
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,4,4,power_law_1.01,0.027324159145355225
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,4,4,power_law_1.01,0.03688960075378418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,4,4,power_law_1.01,0.9436608123779298
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,4,4,power_law_1.01,0.04198016166687012
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,4,4,power_law_1.01,0.04634880065917969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,4,4,power_law_1.01,0.05003136157989502
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,4,4,power_law_1.01,1.2410137939453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,4,4,power_law_1.01,0.06263167858123779
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,4,4,power_law_1.01,0.07255296230316162
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,4,4,power_law_1.01,0.07584896087646484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,4,4,power_law_1.01,0.07971583843231202
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,4,4,power_law_1.01,0.07665919780731201
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,4,4,power_law_1.01,0.07856512069702148
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,4,4,power_law_1.01,1.7883532714843748
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,4,4,power_law_1.01,0.0865395164489746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,4,4,power_law_1.01,0.09283455848693847
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,4,4,power_law_1.01,0.1002393627166748
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,4,4,power_law_1.01,0.11997568130493164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,4,4,power_law_1.01,0.08193920135498048
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,4,4,power_law_1.01,2.4219981384277345
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,4,4,power_law_1.01,0.08425215721130372
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,4,4,power_law_1.01,0.18833152770996092
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,4,4,power_law_1.01,0.08368000030517578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,4,4,power_law_1.01,0.21535360336303713
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,4,4,power_law_1.01,0.3114303970336914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,4,4,power_law_1.01,0.3883033752441406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,4,4,power_law_1.2,0.046092801094055176
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,4,4,power_law_1.2,0.04835328102111817
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,4,4,power_law_1.01,0.6910489654541015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,4,4,power_law_1.01,0.1402239990234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,4,4,power_law_1.2,0.05512447834014893
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,4,4,power_law_1.2,0.06258560180664062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,4,4,power_law_1.01,1.0080345916748048
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,4,4,power_law_1.2,0.08232192039489747
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,4,4,power_law_1.2,0.11851391792297364
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,4,4,power_law_1.2,0.12209024429321289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,4,4,power_law_1.2,0.1288102436065674
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,4,4,power_law_1.2,0.133372163772583
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,4,4,power_law_1.01,0.536126708984375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,4,4,power_law_1.2,0.12800895690917968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,4,4,power_law_1.2,0.13058303833007812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,4,4,power_law_1.2,0.13757439613342287
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,4,4,power_law_1.2,0.14167679786682127
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,4,4,power_law_1.2,0.14550399780273438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,4,4,power_law_1.2,0.15690624237060546
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,4,4,power_law_1.2,0.06498816013336181
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,4,4,power_law_1.2,0.19208192825317383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,4,4,power_law_1.2,0.2314931106567383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,4,4,power_law_1.2,0.31427328109741215
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,4,4,power_law_1.01,1.338067169189453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,4,4,power_law_1.2,0.13426048278808594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,4,4,power_law_1.2,0.39338241577148436
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,4,4,power_law_1.2,0.7188761901855469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,4,4,power_law_1.2,0.5399449539184571
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,4,4,power_law_1.2,0.028829441070556638
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,4,4,power_law_1.2,1.0363455963134764
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,4,4,power_law_1.2,0.03436800003051758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,4,4,power_law_1.2,0.04148096084594727
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,4,4,power_law_1.2,0.045895681381225586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,4,4,power_law_1.2,0.04840320110321045
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,4,4,power_law_1.2,0.06342912197113038
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,4,4,power_law_1.2,0.07297920227050782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,4,4,power_law_1.2,0.07588096141815186
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,4,4,power_law_1.2,0.07874176025390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,4,4,power_law_1.2,0.0844979190826416
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,4,4,power_law_1.2,0.08457088470458984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,4,4,power_law_1.2,0.07962751865386963
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,4,4,power_law_1.2,0.08111743927001953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,4,4,power_law_1.2,0.0848102378845215
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,4,4,power_law_1.2,0.08875776290893554
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,4,4,power_law_1.2,0.09217535972595214
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,4,4,power_law_1.2,0.10298879623413086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,4,4,power_law_1.2,0.12037247657775878
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,4,4,power_law_1.2,0.15537919998168945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,4,4,power_law_1.2,0.18598783493041993
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,4,4,power_law_1.2,0.22424320220947264
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,4,4,power_law_1.2,0.3125004768371582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,4,4,power_law_1.2,0.38540542602539063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,1,16,power_law_1.2,4.100660400390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,4,4,power_law_1.2,2.049381103515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,8,2,balanced,0.024177920818328858
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,8,2,balanced,0.03873023986816406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,8,2,balanced,0.047343358993530274
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,8,2,balanced,0.06133120059967041
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,4,4,power_law_1.2,0.5428377532958985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,8,2,balanced,0.09100671768188476
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,8,2,balanced,0.09144448280334473
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,8,2,balanced,0.09360511779785156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,8,2,balanced,0.0945958423614502
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,8,2,balanced,0.12348544120788574
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,8,2,balanced,0.12410112380981446
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,8,2,balanced,0.12553728103637696
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,8,2,balanced,0.12740351676940917
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,4,4,power_law_1.2,1.198040313720703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,8,2,balanced,0.13027071952819824
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,8,2,balanced,0.13246208190917969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,8,2,balanced,0.13533439636230468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,8,2,balanced,0.13947263717651368
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,8,2,balanced,0.1463270378112793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,8,2,balanced,0.161975040435791
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,8,2,balanced,0.17723520278930666
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,8,2,balanced,0.23887872695922852
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,8,2,balanced,0.2865817642211914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,8,2,balanced,0.4088409423828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,8,2,balanced,0.5212710571289063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,8,2,balanced,0.7552601623535156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,8,2,balanced,0.9701516723632813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,4,4,power_law_1.2,0.7109452819824218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,8,2,balanced,0.02690943956375122
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,8,2,balanced,1.4348825073242186
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,8,2,balanced,0.03431423902511597
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,8,2,balanced,0.04969344139099121
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,8,2,balanced,0.07605120182037353
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,4,4,power_law_1.2,1.0559859466552735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,8,2,balanced,0.07735551834106445
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,8,2,balanced,0.07711872100830078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,8,2,balanced,0.07817855834960938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,8,2,balanced,0.09360896110534668
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,8,2,balanced,0.09507712364196777
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,8,2,balanced,0.09650303840637206
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,8,2,balanced,0.09899904251098632
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,8,2,balanced,0.10025343894958497
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,8,2,balanced,0.10355711936950684
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,8,2,balanced,0.10611712455749511
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,8,2,balanced,0.10896384239196777
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,8,2,balanced,0.1147430419921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,8,2,balanced,0.12747391700744629
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,8,2,balanced,0.1417088031768799
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,8,2,balanced,0.16889215469360352
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,8,2,balanced,0.1969228744506836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,8,2,balanced,0.29781375885009764
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,8,2,balanced,0.3502937698364258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,8,2,balanced,1.9156953430175783
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,8,2,balanced,0.5123673629760742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,8,2,power_law_1.01,0.02983167886734009
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,8,2,balanced,0.6734770965576172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,8,2,power_law_1.01,0.0432908821105957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,8,2,power_law_1.01,0.050705919265747065
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,8,2,balanced,0.9899263763427735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,8,2,power_law_1.01,0.05537792205810547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,8,2,power_law_1.01,0.062447361946105964
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,8,2,power_law_1.01,0.08490752220153809
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,8,2,power_law_1.01,0.07961472034454345
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,8,2,balanced,1.3120346069335938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,8,2,power_law_1.01,0.08660096168518067
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,8,2,power_law_1.01,0.11608192443847656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,8,2,power_law_1.01,0.11829119682312013
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,8,2,power_law_1.01,0.1237440013885498
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,8,2,power_law_1.01,0.12088319778442383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,8,2,power_law_1.01,0.12618495941162108
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,8,2,power_law_1.01,0.13157888412475588
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,8,2,power_law_1.01,0.13660544395446778
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,8,2,power_law_1.01,0.1402726364135742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,8,2,power_law_1.01,0.14819328308105467
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,8,2,power_law_1.01,0.1692812728881836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,8,2,power_law_1.01,0.1903219223022461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,8,2,power_law_1.01,0.2722559928894043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,8,2,balanced,0.02506623983383179
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,8,2,power_law_1.01,0.3393830490112305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,8,2,power_law_1.01,0.4724070358276367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,8,2,power_law_1.01,0.5770841598510742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,8,2,power_law_1.01,0.8095577239990235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,8,2,power_law_1.01,0.026206719875335693
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,8,2,power_law_1.01,1.0784448242187499
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,8,2,power_law_1.01,0.033397760391235355
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,8,2,power_law_1.01,0.04027647972106933
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,8,2,power_law_1.01,0.04506368160247803
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,8,2,power_law_1.01,0.052925438880920404
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,8,2,power_law_1.01,1.6337625122070314
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,8,2,power_law_1.01,0.06747647762298584
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,8,2,power_law_1.01,0.07031551837921143
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,8,2,power_law_1.01,0.07256447792053222
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,8,2,power_law_1.01,0.08621055603027343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,8,2,power_law_1.01,0.08854528427124023
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,8,2,power_law_1.01,0.09177215576171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,8,2,power_law_1.01,0.09516544342041015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,8,2,power_law_1.01,0.09681023597717285
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,8,2,power_law_1.01,0.10217215538024901
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,8,2,power_law_1.01,0.10526592254638671
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,8,2,power_law_1.01,0.11356800079345704
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,8,2,power_law_1.01,0.12676480293273926
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,8,2,power_law_1.01,0.15127167701721192
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,8,2,power_law_1.01,0.16936832427978515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,8,2,power_law_1.01,0.20731136322021487
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,8,2,power_law_1.01,0.23806592941284183
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,8,2,power_law_1.01,0.3340825653076172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,8,2,power_law_1.01,0.4068876647949219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,8,2,power_law_1.01,2.096600341796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,8,2,power_law_1.01,0.5886937713623046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,8,2,power_law_1.2,0.029516799449920656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,8,2,power_law_1.01,0.74648193359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,8,2,power_law_1.2,0.04214144229888916
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,8,2,power_law_1.2,0.04918272018432617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,8,2,power_law_1.2,0.05177984237670898
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,8,2,power_law_1.01,1.0820684814453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,8,2,power_law_1.2,0.08070143699645996
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,8,2,power_law_1.2,0.08518527984619141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,8,2,power_law_1.2,0.0881612777709961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,8,2,power_law_1.2,0.11875967979431153
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,8,2,power_law_1.2,0.12131327629089354
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,8,2,power_law_1.01,1.4581939697265625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,8,2,power_law_1.2,0.12397439956665039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,8,2,power_law_1.2,0.12552831649780274
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,8,2,power_law_1.2,0.12934016227722167
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,8,2,power_law_1.2,0.1330675220489502
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,8,2,power_law_1.2,0.1386521625518799
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,8,2,power_law_1.2,0.14270976066589355
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,8,2,power_law_1.2,0.15264384269714354
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,8,2,power_law_1.2,0.17242752075195314
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,8,2,power_law_1.2,0.20039039611816406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,8,2,power_law_1.2,0.27983488082885744
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,8,2,power_law_1.2,0.35608959197998047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,8,2,power_law_1.2,0.6077836990356446
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,8,2,power_law_1.2,0.47840641021728514
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,8,2,power_law_1.2,0.8666073608398438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,8,2,power_law_1.2,0.059182081222534176
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,2,8,power_law_1.2,1.5760397338867187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,8,2,power_law_1.2,0.026225919723510745
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,8,2,power_law_1.2,0.03387775897979736
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,8,2,power_law_1.2,0.03885440111160278
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,8,2,power_law_1.2,0.04163072109222412
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,8,2,power_law_1.2,0.04844927787780762
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,8,2,power_law_1.2,0.07184256076812744
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,8,2,power_law_1.2,0.0702233600616455
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,8,2,power_law_1.2,0.07260672092437745
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,8,2,power_law_1.2,0.08839296340942383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,8,2,power_law_1.2,0.09178367614746094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,8,2,power_law_1.2,0.09409407615661622
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,8,2,power_law_1.2,0.09637375831604003
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,8,2,power_law_1.2,0.0988479995727539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,8,2,power_law_1.2,0.10236800193786622
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,8,2,power_law_1.2,0.11078271865844727
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,8,2,power_law_1.2,0.12512127876281737
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,8,2,power_law_1.2,0.13603455543518067
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,8,2,power_law_1.2,0.15335424423217775
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,8,2,power_law_1.2,0.1727244758605957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,8,2,power_law_1.2,0.20726015090942385
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,8,2,power_law_1.2,0.2413043212890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,8,2,power_law_1.2,0.3406272125244141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,8,2,power_law_1.2,0.41995903015136715
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,8,2,power_law_1.2,1.704521026611328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,8,2,power_law_1.2,0.5926950454711915
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,16,1,balanced,0.03085184097290039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,8,2,power_law_1.2,1.124042205810547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,8,2,power_law_1.2,0.7498086547851562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,16,1,balanced,0.04808576107025146
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,8,2,power_law_1.2,1.1097484588623048
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,16,1,balanced,0.11673664093017577
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,16,1,balanced,0.059163517951965325
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,16,1,balanced,0.11729215621948241
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,16,1,balanced,0.11725440025329589
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,16,1,balanced,0.12144512176513671
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,16,1,balanced,0.0763756799697876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,16,1,balanced,0.12252160072326661
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,16,1,balanced,0.14301247596740724
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,16,1,balanced,0.1442905616760254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,16,1,balanced,0.11876031875610352
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,16,1,balanced,0.14698368072509765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,16,1,balanced,0.1490176010131836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,16,1,balanced,0.1516217613220215
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,16,1,balanced,0.15714559555053711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,16,1,balanced,0.17038463592529296
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,16,1,balanced,0.18410688400268554
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,16,1,balanced,0.24352447509765626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,16,1,balanced,0.29631872177124025
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,16,1,balanced,0.4134624099731445
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,16,1,balanced,0.505470085144043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,16,1,balanced,0.12691519737243653
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,16,1,balanced,0.760601577758789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,16,1,balanced,0.03253119945526123
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,16,1,balanced,0.0368287992477417
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,16,1,balanced,1.0003910064697266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,16,1,balanced,0.047793917655944824
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,16,1,balanced,0.06839424133300781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,16,1,balanced,0.10700544357299804
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,16,1,balanced,0.10755840301513672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,16,1,balanced,0.10855104446411132
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,16,1,balanced,0.10980992317199707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,16,1,balanced,0.11081983566284179
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,16,1,balanced,0.11219648361206054
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,4,4,power_law_1.2,2.642581787109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,16,1,balanced,0.11339776039123535
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,16,1,balanced,0.1343552017211914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,16,1,balanced,0.13577280044555665
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,16,1,balanced,1.9936621093749998
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,16,1,balanced,0.1382649612426758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,16,1,balanced,0.13989503860473634
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,16,1,balanced,1.4635507202148437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,16,1,balanced,0.1405241584777832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,16,1,balanced,0.14652992248535157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,16,1,balanced,0.15813823699951174
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,16,1,balanced,0.17001279830932617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,16,1,balanced,0.194705924987793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,16,1,balanced,0.21714239120483397
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,16,1,balanced,0.33840766906738284
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,16,1,balanced,0.38464256286621096
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,16,1,balanced,0.5553926467895508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,16,1,balanced,0.7339718627929688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,16,1,power_law_1.01,0.0493120002746582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,16,1,power_law_1.01,0.059134721755981445
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,16,1,balanced,1.080843505859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,16,1,power_law_1.01,0.06886975765228272
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,16,1,power_law_1.01,0.08359168052673341
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,16,1,power_law_1.01,0.10840319633483886
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,16,1,balanced,1.4392735290527345
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,16,1,power_law_1.01,0.1111961555480957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,16,1,power_law_1.01,0.11648320198059081
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,16,1,power_law_1.01,0.11988608360290527
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,16,1,power_law_1.01,0.12891263961791993
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,16,1,power_law_1.01,0.03069823980331421
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,16,1,power_law_1.01,0.1396070384979248
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,16,1,power_law_1.01,0.14190591812133788
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,16,1,power_law_1.01,0.14667712211608888
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,16,1,power_law_1.01,0.15142080307006836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,16,1,power_law_1.01,0.15459391593933106
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,16,1,power_law_1.01,0.16368255615234376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,16,1,power_law_1.01,0.18255935668945314
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,16,1,power_law_1.01,0.20078271865844727
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,16,1,power_law_1.01,0.26008575439453124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,4,4,power_law_1.2,1.4427084350585937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,16,1,power_law_1.01,0.10350975990295411
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,16,1,power_law_1.01,0.320118408203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,16,1,power_law_1.01,0.45110591888427737
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,16,1,power_law_1.01,0.5604627227783203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,16,1,power_law_1.01,0.032020480632781984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,16,1,power_law_1.01,0.7875660705566406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,16,1,power_law_1.01,0.039698560237884525
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,16,1,power_law_1.01,0.0497651195526123
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,16,1,power_law_1.01,0.06044991970062256
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,16,1,power_law_1.01,0.07293888092041015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,16,1,power_law_1.01,1.0210546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,16,1,power_law_1.01,0.09782272338867187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,16,1,power_law_1.01,0.09926976203918457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,16,1,power_law_1.01,0.10154815673828126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,16,1,power_law_1.01,0.1046784019470215
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,16,1,power_law_1.01,0.10519488334655762
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,16,1,power_law_1.01,0.1092198371887207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,16,1,power_law_1.01,1.5079866027832032
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,16,1,power_law_1.01,0.13201024055480956
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,16,1,power_law_1.01,0.13826239585876465
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,16,1,power_law_1.01,0.14280447959899903
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,16,1,power_law_1.01,0.14844160079956054
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,16,1,power_law_1.01,0.1660166358947754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,16,1,power_law_1.01,1.9767674255371095
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,16,1,power_law_1.01,0.18808256149291994
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,16,1,power_law_1.01,0.20638784408569336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,16,1,power_law_1.01,0.24193279266357423
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,16,1,power_law_1.01,0.2836704063415527
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,16,1,power_law_1.01,0.13054719924926758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,16,1,power_law_1.01,0.3816633605957031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,16,1,power_law_1.01,0.45437183380126955
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,16,1,power_law_1.01,0.6283091354370117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,16,1,power_law_1.01,0.8025222778320312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,16,1,power_law_1.2,0.04936063766479492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,16,1,power_law_1.2,0.0660371208190918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,16,1,power_law_1.01,1.1580786895751953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,16,1,power_law_1.2,0.07913472175598144
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,16,1,power_law_1.01,1.5098074340820313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,16,1,power_law_1.2,0.1093945598602295
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,16,1,power_law_1.2,0.031024639606475828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,16,1,power_law_1.2,0.11354687690734863
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,16,1,power_law_1.2,0.11872511863708497
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,16,1,power_law_1.2,0.12110912322998046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,16,1,power_law_1.2,0.12956031799316406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,16,1,power_law_1.2,0.14343104362487794
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,16,1,power_law_1.2,0.14900416374206543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,16,1,power_law_1.2,0.05818624019622802
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,16,1,power_law_1.2,0.1534118366241455
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,16,1,power_law_1.2,0.1560921573638916
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,16,1,power_law_1.2,0.16577983856201173
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,16,1,power_law_1.2,0.1847327995300293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,16,1,power_law_1.2,0.2046886444091797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,16,1,power_law_1.2,0.1044281578063965
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,16,1,power_law_1.2,0.2676435279846191
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,16,1,power_law_1.2,0.14118911743164062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,16,1,power_law_1.2,0.32075134277343753
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,16,1,power_law_1.2,0.4572371292114258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,16,1,power_law_1.2,0.5643231964111328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,16,1,power_law_1.2,0.039888639450073246
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,16,1,power_law_1.2,0.048602237701416015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,16,1,power_law_1.2,0.058167037963867185
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,16,1,power_law_1.2,0.06920063972473145
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,16,1,power_law_1.2,0.0961184024810791
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,16,1,power_law_1.2,0.09966272354125977
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,16,1,power_law_1.2,0.03202431917190552
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,16,1,power_law_1.2,0.7924288177490235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,16,1,power_law_1.2,0.10585791587829589
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,16,1,power_law_1.2,0.10629183769226073
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,16,1,power_law_1.2,0.132161922454834
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,16,1,power_law_1.2,0.10334464073181153
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,16,1,power_law_1.2,0.13431936264038086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,16,1,power_law_1.2,0.1408403205871582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,16,1,power_law_1.2,0.10889792442321777
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,16,1,power_law_1.2,0.1557209587097168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,16,1,power_law_1.2,0.1756460762023926
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,16,1,power_law_1.2,0.1913043212890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,16,1,power_law_1.2,0.20870784759521482
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,16,1,power_law_1.2,0.24915327072143553
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,16,1,power_law_1.2,0.14802176475524903
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,16,1,power_law_1.2,0.2860166358947754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,8,2,power_law_1.2,2.200966339111328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,16,1,power_law_1.2,0.3821113586425781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,16,1,power_law_1.2,1.0276525115966797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,16,1,power_law_1.2,0.46014846801757814
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,1,16,balanced,0.06758272171020507
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,1,16,balanced,0.06746496200561523
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,1,16,balanced,0.07869952201843262
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,1,16,balanced,0.1116979217529297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,1,16,balanced,0.20960767745971678
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,1,16,balanced,0.3515046310424805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,1,16,balanced,0.3544243240356445
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,1,16,balanced,0.35965312957763673
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,1,16,balanced,0.36620033264160157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,1,16,balanced,0.3715238571166992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,16,1,power_law_1.2,0.6344691085815429
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,1,16,balanced,0.37561344146728515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,1,16,balanced,0.3792063903808594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,8,2,power_law_1.2,1.4806285095214844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,1,16,balanced,0.3856819152832031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,1,16,balanced,0.38189952850341796
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,1,16,balanced,0.39080448150634767
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,1,16,balanced,0.40421504974365235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,1,16,balanced,0.4245068740844727
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,1,16,balanced,0.4543936157226563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,1,16,balanced,0.4886105728149414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,1,16,balanced,0.6091212844848632
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,1,16,balanced,0.6420979309082031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,1,16,balanced,0.8668006134033203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,1,16,balanced,1.1019110107421874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,1,16,balanced,1.4403724670410156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,1,16,balanced,1.9141580200195314
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,1,16,balanced,0.04140927791595459
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,1,16,balanced,0.04145792007446289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,16,1,power_law_1.2,0.8079948425292969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,1,16,balanced,0.06802815914154052
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,1,16,balanced,2.7520050048828124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,1,16,balanced,0.12359935760498046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,1,16,balanced,0.15080063819885253
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,1,16,balanced,0.15216511726379395
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,1,16,balanced,0.15691519737243653
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,1,16,balanced,0.157009916305542
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,1,16,balanced,0.1597708797454834
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,1,16,balanced,0.16387584686279297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,1,16,balanced,0.16717824935913086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,1,16,balanced,0.1711334419250488
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,1,16,balanced,0.16712703704833984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,1,16,balanced,3.6259393310546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,1,16,balanced,0.16971263885498047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,1,16,balanced,0.17533567428588867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,1,16,balanced,0.18567935943603514
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,1,16,balanced,0.20047359466552733
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,1,16,balanced,0.21623680114746097
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,1,16,balanced,0.24727935791015626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,1,16,balanced,0.2725734329223633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,1,16,balanced,0.39377792358398434
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,1,16,balanced,0.45220481872558593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,1,16,balanced,0.04953343868255615
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,1,16,balanced,0.6162982559204101
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,1,16,balanced,0.8108364868164063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,1,16,balanced,1.0931750488281249
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,1,16,power_law_1.01,0.19123199462890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,1,16,power_law_1.01,0.2762393569946289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,1,16,power_law_1.01,0.2766412734985352
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,1,16,balanced,1.4500416564941407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,1,16,power_law_1.01,0.2965683174133301
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,1,16,power_law_1.01,0.2842521667480469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,1,16,power_law_1.01,0.31159551620483394
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,1,16,power_law_1.01,0.32333438873291015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,1,16,power_law_1.01,0.3050777626037598
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,1,16,power_law_1.01,0.3354367828369141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,1,16,power_law_1.01,0.37103359222412113
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,1,16,power_law_1.01,0.3552473449707031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,1,16,power_law_1.01,0.3636646270751953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,1,16,power_law_1.01,0.36919551849365234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,1,16,power_law_1.01,0.3852889633178711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,1,16,power_law_1.01,0.42710399627685547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,1,16,power_law_1.01,0.4594099044799805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,1,16,power_law_1.01,0.48283264160156253
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,1,16,power_law_1.01,0.5391078567504882
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,1,16,power_law_1.01,0.601580810546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,16,1,power_law_1.2,1.5113127136230469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,1,16,power_law_1.01,0.8584832000732423
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,1,16,power_law_1.01,1.0117964935302735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,1,16,power_law_1.01,1.479436798095703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,1,16,power_law_1.01,2.038899230957031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,1,16,power_law_1.01,0.12470656394958497
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,1,16,power_law_1.01,0.13659135818481444
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,1,16,power_law_1.01,2.65235595703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,1,16,power_law_1.01,0.12235903739929199
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,1,16,power_law_1.01,0.13620608329772949
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,1,16,power_law_1.01,0.10970879554748536
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,1,16,power_law_1.01,0.14639103889465332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,1,16,power_law_1.01,0.14841216087341308
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,1,16,power_law_1.01,3.7394534301757814
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,1,16,power_law_1.01,0.1537395191192627
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,1,16,power_law_1.01,0.15545727729797362
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,1,16,power_law_1.01,0.16170495986938477
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,1,16,power_law_1.01,0.16418048858642578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,1,16,power_law_1.01,0.168035831451416
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,1,16,power_law_1.01,0.16707456588745118
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,1,16,power_law_1.01,0.16724224090576173
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,1,16,power_law_1.01,0.1768307113647461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,1,16,power_law_1.01,0.18516223907470702
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,1,16,power_law_1.01,0.19618175506591798
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,1,16,power_law_1.01,0.21857791900634765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,1,16,power_law_1.01,0.2713254356384277
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,1,16,power_law_1.01,0.32915328979492187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,1,16,power_law_1.01,0.39246337890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,1,16,power_law_1.01,5.8310876464843755
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,1,16,power_law_1.01,0.5378931045532227
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,1,16,power_law_1.01,0.7661977386474609
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,1,16,power_law_1.01,0.9651251220703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,1,16,power_law_1.2,0.19414655685424803
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,1,16,power_law_1.01,1.3303053283691406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,1,16,power_law_1.2,0.2089459228515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,1,16,power_law_1.01,7.2474255371093745
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,1,16,power_law_1.2,0.26197248458862304
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,1,16,power_law_1.01,1.9282521057128907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,1,16,power_law_1.2,0.27724288940429687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,1,16,power_law_1.2,0.24167552947998047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,1,16,power_law_1.2,0.3211494445800781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,1,16,power_law_1.2,0.31611263275146484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,1,16,power_law_1.2,0.32252288818359376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,1,16,power_law_1.2,0.33662208557128903
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,1,16,power_law_1.2,0.3502694320678711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,1,16,power_law_1.2,0.3584320068359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,1,16,power_law_1.2,0.38692096710205076
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,1,16,power_law_1.2,0.3995238494873047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,1,16,power_law_1.01,2.89918212890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,1,16,power_law_1.2,0.4931327819824219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,1,16,power_law_1.2,0.4997491073608399
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,16,1,power_law_1.2,1.1648786926269532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,1,16,power_law_1.2,0.5507622528076171
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,1,16,power_law_1.2,0.6153049468994141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,1,16,power_law_1.2,0.6947917175292969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,1,16,power_law_1.2,0.42756095886230466
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,1,16,power_law_1.2,0.9323571014404297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,1,16,power_law_1.2,1.1572045135498048
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,1,16,power_law_1.2,1.6669235229492188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,1,16,power_law_1.2,0.12600831985473632
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,1,16,power_law_1.2,2.1973951721191405
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,1,16,power_law_1.2,0.11449088096618651
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,1,16,power_law_1.2,0.13873791694641113
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,1,16,power_law_1.2,0.13748096466064452
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,1,16,power_law_1.2,0.11020416259765624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,1,16,power_law_1.2,0.14513152122497558
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,1,16,power_law_1.2,0.14773119926452635
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,1,16,power_law_1.2,3.2189901733398436
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,1,16,power_law_1.2,0.15508480072021485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,1,16,power_law_1.2,0.15987071990966797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,1,16,power_law_1.2,0.16249727249145507
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,1,16,power_law_1.2,0.17277055740356445
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,1,16,power_law_1.2,0.17310592651367188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,1,16,power_law_1.2,0.15194751739501952
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,1,16,power_law_1.2,4.454379577636719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,1,16,power_law_1.2,0.1730726432800293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,1,16,power_law_1.2,0.1834739112854004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,1,16,power_law_1.2,0.22072063446044923
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,1,16,power_law_1.2,0.2501734352111816
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,1,16,power_law_1.2,0.2835264015197754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,1,16,power_law_1.2,0.3510259246826172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,1,16,power_law_1.2,0.439552001953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,1,16,power_law_1.2,0.18941951751708985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,1,16,power_law_1.2,0.6431231689453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,1,16,power_law_1.2,0.8983679962158202
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,1,16,power_law_1.2,7.461830444335938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,2,8,balanced,0.05911808013916016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,2,8,balanced,0.06244863986968994
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,1,16,power_law_1.2,1.2125170898437498
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,2,8,balanced,0.07703423976898194
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,2,8,balanced,0.11323007583618164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,2,8,balanced,0.2123788833618164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,2,8,balanced,0.34293888092041014
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,2,8,balanced,0.34861183166503906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,2,8,balanced,0.35766273498535156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,2,8,balanced,0.36407550811767575
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,1,16,power_law_1.2,1.8386802673339844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,2,8,balanced,0.3706700897216797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,2,8,balanced,0.36580734252929686
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,2,8,balanced,0.38296958923339847
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,2,8,balanced,0.3760255813598633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,2,8,balanced,0.38721920013427735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,2,8,balanced,0.3989299011230469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,2,8,balanced,0.4160921478271485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,2,8,balanced,0.4344960021972656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,2,8,balanced,0.47858943939208987
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,16,1,power_law_1.2,1.9922688293457032
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,2,8,balanced,0.6209395217895508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,2,8,balanced,0.6839564514160157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,2,8,balanced,0.9617612457275391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,2,8,balanced,0.516677131652832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,2,8,balanced,1.1943052673339845
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,2,8,balanced,1.623937225341797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,2,8,balanced,0.03265023946762085
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,2,8,balanced,0.03845119953155517
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,2,8,balanced,0.04891520023345947
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,2,8,balanced,0.07153024196624755
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,2,8,balanced,0.1262387180328369
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,2,8,balanced,0.1576371192932129
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,2,8,balanced,0.1627916717529297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,2,8,balanced,3.118262939453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,2,8,balanced,2.1631436157226562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,2,8,balanced,0.16858240127563479
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,2,8,balanced,0.1714227294921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,1,16,power_law_1.2,8.36247802734375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,2,8,balanced,0.16407424926757813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,2,8,balanced,0.16674688339233398
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,2,8,balanced,0.17510528564453126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,2,8,balanced,0.1698111915588379
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,2,8,balanced,4.066947937011719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,2,8,balanced,0.17942144393920897
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,2,8,balanced,0.1878438377380371
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,2,8,balanced,0.1657881546020508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,2,8,balanced,0.20585599899291993
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,2,8,balanced,0.23071359634399413
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,2,8,balanced,0.2960755157470703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,2,8,balanced,0.3435251235961914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,2,8,balanced,0.5050099182128907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,2,8,balanced,0.6041664123535156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,2,8,balanced,0.8357708740234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,2,8,balanced,0.2501414489746094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,1,16,power_law_1.2,2.8055438232421874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,2,8,balanced,1.100175323486328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,2,8,power_law_1.01,0.10880512237548827
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,2,8,power_law_1.01,0.16314495086669922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,2,8,power_law_1.01,0.14819199562072755
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,2,8,power_law_1.01,0.19490304946899414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,2,8,balanced,1.5227813720703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,2,8,power_law_1.01,0.31388671874999996
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,2,8,power_law_1.01,0.3088307189941406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,2,8,power_law_1.01,0.3183692741394043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,2,8,power_law_1.01,0.3489945602416992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,16,1,power_law_1.2,1.5160671997070314
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,2,8,power_law_1.01,0.3630681610107422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,2,8,power_law_1.01,0.3571097564697266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,2,8,power_law_1.01,0.3723737716674805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,2,8,power_law_1.01,0.19966592788696289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,2,8,power_law_1.01,0.3949248123168945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,2,8,power_law_1.01,0.4287577438354492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,2,8,power_law_1.01,0.4652556610107422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,2,8,power_law_1.01,0.5367628860473632
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,1,16,power_law_1.2,3.408572082519531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,2,8,power_law_1.01,0.5725734329223633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,2,8,power_law_1.01,0.6951283264160156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,2,8,power_law_1.01,0.39693054199218747
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,2,8,power_law_1.01,0.8479961395263672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,2,8,power_law_1.01,1.1152537536621092
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,2,8,power_law_1.01,0.31327999114990235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,2,8,balanced,2.0239500427246093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,2,8,power_law_1.01,0.07106048107147217
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,2,8,power_law_1.01,0.10601984024047853
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,2,8,power_law_1.01,0.09535743713378905
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,2,8,power_law_1.01,0.125927677154541
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,2,8,power_law_1.01,0.11635711669921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,2,8,power_law_1.01,0.15500543594360353
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,2,8,power_law_1.01,2.139737548828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,2,8,power_law_1.01,0.1530444812774658
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,2,8,power_law_1.01,0.15802751541137697
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,2,8,power_law_1.01,0.16509824752807617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,2,8,power_law_1.01,2.5335705566406252
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,2,8,power_law_1.01,0.1604902458190918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,2,8,power_law_1.01,0.16741376876831054
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,2,8,power_law_1.01,0.16540159225463866
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,2,8,power_law_1.01,0.16544639587402343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,2,8,power_law_1.01,0.1625356864929199
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,2,8,power_law_1.01,0.1710207939147949
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,2,8,power_law_1.01,0.17729663848876953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,2,8,power_law_1.01,0.19314687728881835
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,2,8,power_law_1.01,0.22476032257080075
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,2,8,power_law_1.01,0.2662348747253418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,2,8,power_law_1.01,0.3160819244384766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,2,8,power_law_1.01,0.3764748764038086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,2,8,power_law_1.01,1.597015075683594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,2,8,power_law_1.01,0.48717823028564455
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,2,8,power_law_1.01,0.6741222381591797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,2,8,power_law_1.2,0.10851200103759766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,2,8,power_law_1.01,0.9200371551513673
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,2,8,power_law_1.2,0.12117759704589844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,2,8,power_law_1.01,4.5456781005859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,2,8,power_law_1.2,0.1623027229309082
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,2,8,power_law_1.2,0.15040639877319334
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,2,8,power_law_1.01,1.2185894775390627
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,2,8,power_law_1.2,0.3113113594055176
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,2,8,power_law_1.2,0.2943129539489746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,2,8,power_law_1.2,0.323559684753418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,2,8,power_law_1.01,5.419976806640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,2,8,power_law_1.2,0.3161280059814453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,2,8,power_law_1.2,0.3486374282836914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,2,8,power_law_1.01,1.789213409423828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,2,8,power_law_1.2,0.36561279296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,2,8,power_law_1.2,0.37743999481201174
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,2,8,power_law_1.2,0.4029977416992187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,2,8,power_law_1.2,0.4176076889038086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,2,8,power_law_1.01,2.356400604248047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,2,8,power_law_1.2,0.5183987045288087
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,2,8,power_law_1.2,0.5603379058837891
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,2,8,power_law_1.2,0.6061260986328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,2,8,power_law_1.2,0.3615961456298828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,2,8,power_law_1.2,0.7160806274414062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,2,8,power_law_1.2,0.20741247177124023
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,2,8,power_law_1.2,0.9261043548583985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,2,8,power_law_1.2,0.44661376953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,2,8,power_law_1.2,0.07013504028320312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,2,8,power_law_1.2,0.07170688152313233
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,2,8,power_law_1.2,0.0934272003173828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,2,8,power_law_1.2,0.12427647590637207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,2,8,power_law_1.2,1.5961932373046874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,2,8,power_law_1.2,0.08906496047973633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,2,8,power_law_1.2,0.14819071769714356
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,2,8,power_law_1.2,0.15670656204223632
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,2,8,power_law_1.2,0.15674880027770996
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,2,8,power_law_1.2,0.16067712783813476
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,2,8,power_law_1.2,0.16266239166259766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,2,8,power_law_1.2,0.1676710319519043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,2,8,power_law_1.2,0.16360319137573243
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,2,8,power_law_1.2,0.16481920242309572
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,2,8,power_law_1.2,2.7328564453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,2,8,power_law_1.2,0.16972288131713867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,2,8,power_law_1.2,0.17460224151611328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,2,8,power_law_1.2,0.1821721649169922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,2,8,power_law_1.2,0.1944883155822754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,2,8,power_law_1.2,0.23773056030273437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,2,8,power_law_1.2,3.1584933471679686
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,2,8,power_law_1.2,0.27041791915893554
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,2,8,power_law_1.2,1.3105523681640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,2,8,power_law_1.2,0.393809928894043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,2,8,power_law_1.2,0.5370790481567382
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,2,8,power_law_1.2,0.6736434936523438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,4,4,balanced,0.052110080718994145
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,4,4,balanced,0.06119040012359619
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,2,8,power_law_1.2,0.31872512817382814
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,4,4,balanced,0.07674496173858643
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,2,8,power_law_1.2,0.9648793792724609
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,2,8,power_law_1.2,1.2847436523437499
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,4,4,balanced,0.1131328010559082
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,4,4,balanced,0.19669248580932616
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,4,4,balanced,0.2649996757507324
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,4,4,balanced,0.34004993438720704
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,4,4,balanced,0.3442943954467773
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,4,4,balanced,0.3459596633911133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,4,4,balanced,0.34680191040039066
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,4,4,balanced,0.35621631622314454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,4,4,balanced,0.3633049774169922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,4,4,balanced,0.37790206909179686
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,4,4,balanced,0.38486785888671876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,4,4,balanced,0.40050430297851564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,4,4,balanced,0.34429695129394533
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,4,4,balanced,0.40732799530029296
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,4,4,balanced,0.46152702331542966
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,4,4,balanced,0.4974579238891602
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,4,4,balanced,0.6108633422851563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,4,4,balanced,0.6771532440185547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,4,4,balanced,0.9337677001953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,4,4,balanced,1.176895980834961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,2,8,power_law_1.2,4.8040704345703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,4,4,balanced,1.5836531066894532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,4,4,balanced,0.03145215988159179
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,4,4,balanced,0.040295681953430175
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,2,8,power_law_1.2,2.035335693359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,4,4,balanced,0.052426238059997556
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,4,4,balanced,2.094132537841797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,4,4,balanced,0.07551743984222412
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,4,4,balanced,0.1314252758026123
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,4,4,balanced,0.15127039909362794
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,4,4,balanced,0.1709836769104004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,4,4,balanced,0.17464960098266602
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,4,4,balanced,0.17689599990844726
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,4,4,balanced,0.169300479888916
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,4,4,balanced,0.1720755195617676
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,4,4,balanced,0.17447551727294922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,4,4,balanced,0.17915519714355468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,4,4,balanced,0.18486528396606444
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,4,4,balanced,3.0095297241210934
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,4,4,balanced,0.1914406394958496
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,4,4,balanced,0.20204416275024414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,4,4,balanced,0.2272537612915039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,4,4,balanced,0.24453504562377928
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,4,4,balanced,0.28776063919067385
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,4,4,balanced,0.17114240646362305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,4,4,balanced,0.498131217956543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,4,4,balanced,0.5813401412963868
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,4,4,balanced,3.9218981933593753
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,4,4,balanced,0.8300364685058593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,4,4,power_law_1.01,0.07654399871826172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,4,4,balanced,1.0630963134765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,4,4,power_law_1.01,0.113505277633667
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,4,4,power_law_1.01,0.10164223670959474
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,4,4,balanced,0.33701759338378906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,4,4,power_law_1.01,0.14339455604553222
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,4,4,power_law_1.01,0.23070079803466798
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,4,4,power_law_1.01,0.29309312820434574
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,4,4,balanced,1.4655410766601562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,4,4,power_law_1.01,0.31222656249999997
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,4,4,power_law_1.01,0.3188787269592285
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,4,4,power_law_1.01,0.3280486297607422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,4,4,power_law_1.01,0.31292415618896485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,4,4,power_law_1.01,0.32929534912109376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,4,4,balanced,1.944762878417969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,4,4,power_law_1.01,0.3282624053955078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,4,4,power_law_1.01,0.3646284866333008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,4,4,power_law_1.01,0.1726924705505371
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,4,4,power_law_1.01,0.3807244873046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,4,4,power_law_1.01,0.39852798461914063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,4,4,power_law_1.01,0.4308569717407226
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,4,4,power_law_1.01,0.5055654525756836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,4,4,power_law_1.01,0.5372748947143554
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,4,4,power_law_1.01,0.620835838317871
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,4,4,power_law_1.01,0.7397964477539063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,4,4,power_law_1.01,1.0214579010009766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,4,4,power_law_1.01,0.05126016139984131
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,4,4,power_law_1.01,1.2842726135253906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,2,8,power_law_1.2,2.6312103271484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,4,4,power_law_1.01,0.06659711837768555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,4,4,power_law_1.01,0.07466239929199218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,4,4,power_law_1.01,0.09428607940673828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,4,4,power_law_1.01,1.8432205200195313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,4,4,power_law_1.01,0.1141964817047119
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,4,4,power_law_1.01,0.14402175903320313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,4,4,power_law_1.01,0.16296831130981446
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,4,4,power_law_1.01,0.17177215576171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,4,4,power_law_1.01,0.1659596824645996
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,4,4,power_law_1.01,0.1663372802734375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,4,4,power_law_1.01,0.16990848541259768
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,4,4,power_law_1.01,2.492825622558594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,4,4,power_law_1.01,0.16666240692138673
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,4,4,power_law_1.01,0.16701440811157225
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,4,4,power_law_1.01,0.17469823837280274
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,4,4,power_law_1.01,0.17831039428710938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,4,4,power_law_1.01,0.19133567810058594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,4,4,power_law_1.01,0.197076473236084
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,4,4,power_law_1.01,0.2254207992553711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,4,4,power_law_1.01,0.2499660873413086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,4,4,power_law_1.01,0.32823040008544924
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,4,4,power_law_1.01,0.3768511962890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,4,4,power_law_1.01,0.49872512817382814
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,4,4,power_law_1.01,3.4440869140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,4,4,power_law_1.01,0.6106918334960938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,2,8,power_law_1.2,6.887785034179688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,4,4,power_law_1.2,0.08042112350463868
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,4,4,power_law_1.01,0.8543590545654297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,4,4,power_law_1.2,0.122227201461792
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,4,4,power_law_1.2,0.13719807624816896
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,4,4,power_law_1.01,1.1335667419433595
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,4,4,power_law_1.2,0.23948928833007815
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,4,4,power_law_1.2,0.2864934349060059
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,4,4,power_law_1.2,0.28910207748413086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,4,4,power_law_1.01,4.693697204589844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,4,4,power_law_1.2,0.31126272201538085
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,4,4,power_law_1.01,1.7043878173828126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,4,4,power_law_1.2,0.08742527961730957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,4,4,power_law_1.2,0.3281792068481445
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,4,4,power_law_1.2,0.33726848602294923
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,4,4,power_law_1.2,0.13355648040771484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,4,4,power_law_1.2,0.34113025665283203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,4,4,power_law_1.2,0.36190975189208985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,4,4,power_law_1.01,2.128796234130859
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,4,4,power_law_1.2,0.38712703704833984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,4,4,power_law_1.2,0.3917580795288086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,4,4,power_law_1.2,0.4285631942749023
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,4,4,power_law_1.2,0.4637887954711914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,4,4,power_law_1.2,0.4931468963623047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,4,4,power_law_1.2,0.5529766464233399
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,4,4,power_law_1.2,0.6619123077392578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,4,4,power_law_1.2,0.7725325012207032
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,4,4,power_law_1.2,0.051199998855590824
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,4,4,power_law_1.2,1.10748291015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,4,4,power_law_1.2,0.05729152202606201
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,4,4,power_law_1.2,0.0770201587677002
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,4,4,power_law_1.2,1.4923866271972657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,4,4,power_law_1.2,0.0887667179107666
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,4,4,power_law_1.2,0.09595135688781739
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,4,4,power_law_1.2,1.8700837707519533
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,4,4,power_law_1.2,0.16200191497802735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,4,4,power_law_1.2,0.16340736389160154
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,4,4,power_law_1.2,0.1708799934387207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,4,4,power_law_1.2,0.16924543380737306
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,4,4,power_law_1.2,0.16719232559204103
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,4,4,power_law_1.2,0.16759807586669923
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,4,4,power_law_1.2,0.17257856369018554
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,4,4,power_law_1.2,0.17768064498901368
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,4,4,power_law_1.2,0.18317823410034179
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,4,4,power_law_1.2,0.18679168701171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,4,4,power_law_1.2,0.20125055313110352
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,4,4,power_law_1.2,0.144016637802124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,4,4,power_law_1.2,0.23269376754760743
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,4,4,power_law_1.2,0.2760537528991699
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,4,4,power_law_1.2,2.7068325805664064
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,4,4,power_law_1.2,0.32982784271240234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,4,4,power_law_1.2,0.38695297241210935
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,4,4,power_law_1.2,0.5159065628051758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,4,4,power_law_1.2,0.6532339477539063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,8,2,balanced,0.05048575878143311
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,8,2,balanced,0.05901440143585205
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,8,2,balanced,0.0750105619430542
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,8,2,balanced,0.10924799919128418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,8,2,balanced,0.1940492820739746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,4,4,power_law_1.2,0.9590541076660155
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,8,2,balanced,0.25917184829711915
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,8,2,balanced,0.26478847503662106
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,8,2,balanced,0.2671206474304199
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,4,4,power_law_1.2,1.2179724884033203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,8,2,balanced,0.32947582244873047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,8,2,balanced,0.33438079833984374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,4,4,power_law_1.2,3.6130752563476562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,8,2,balanced,0.33776512145996096
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,8,2,balanced,0.34087806701660156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,8,2,balanced,0.3477849578857422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,8,2,balanced,0.3555481719970703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,8,2,balanced,0.39896705627441403
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,8,2,balanced,0.26880640029907227
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,8,2,balanced,0.43448959350585936
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,8,2,balanced,0.47763839721679685
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,8,2,balanced,0.5946495819091797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,8,2,balanced,0.6506777954101562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,8,2,balanced,0.369354248046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,8,2,balanced,0.9251302337646484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,8,2,balanced,1.1764351654052736
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,8,2,balanced,0.031596798896789555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,8,2,balanced,0.04126848220825195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,8,2,balanced,1.571295928955078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,8,2,balanced,0.05308032035827637
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,8,2,balanced,0.08040960311889649
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,8,2,balanced,0.142807035446167
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,8,2,balanced,0.16691455841064454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,8,2,balanced,2.0758285522460938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,8,2,balanced,0.1665305519104004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,8,2,balanced,0.16723840713500976
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,8,2,balanced,0.1699929618835449
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,8,2,balanced,0.20228223800659179
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,8,2,balanced,0.2036800003051758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,8,2,balanced,0.20699903488159183
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,8,2,balanced,0.20893312454223634
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,8,2,balanced,0.21231231689453126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,8,2,balanced,0.21582847595214844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,8,2,balanced,2.964869079589844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,8,2,balanced,0.2199065589904785
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,8,2,balanced,0.22488319396972659
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,8,2,balanced,0.2485977554321289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,8,2,balanced,0.2728883171081543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,8,2,balanced,0.304956169128418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,8,2,balanced,0.3506304168701172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,8,2,balanced,0.5402905654907226
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,8,2,balanced,3.8512039184570312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,8,2,balanced,0.617886734008789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,4,4,power_law_1.2,1.763377990722656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,8,2,balanced,0.8983424377441406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,8,2,power_law_1.01,0.059005441665649416
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,8,2,power_law_1.01,0.07246463775634766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,8,2,power_law_1.01,0.08334976196289062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,8,2,balanced,1.1748300933837892
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,8,2,power_law_1.01,0.1448140811920166
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,8,2,power_law_1.01,0.2105356788635254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,8,2,power_law_1.01,0.21050880432128904
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,8,2,power_law_1.01,0.2363225555419922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,8,2,balanced,1.61955078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,8,2,power_law_1.01,0.24236671447753907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,8,2,power_law_1.01,0.3080307197570801
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,8,2,power_law_1.01,0.3002739143371582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,8,2,balanced,2.1437196350097656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,4,4,power_law_1.2,4.9529138183593755
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,8,2,power_law_1.01,0.30975744247436526
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,8,2,power_law_1.01,0.32790782928466794
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,8,2,power_law_1.01,0.3436991882324219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,8,2,power_law_1.01,0.3754393768310547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,8,2,power_law_1.01,0.412174072265625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,8,2,power_law_1.01,0.47153278350830075
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,8,2,power_law_1.01,0.11716992378234863
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,8,2,power_law_1.01,0.30974975585937503
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,8,2,power_law_1.01,0.5982937622070312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,8,2,power_law_1.01,0.7030745697021484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,8,2,power_law_1.01,0.9455091094970702
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,8,2,power_law_1.01,1.1895948791503907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,8,2,power_law_1.01,0.041898241043090825
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,8,2,power_law_1.01,0.5261171340942383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,8,2,power_law_1.01,0.05111551761627198
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,8,2,power_law_1.01,0.05895679950714111
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,8,2,power_law_1.01,0.084071683883667
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,8,2,power_law_1.01,1.645745849609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,8,2,power_law_1.01,0.0916659164428711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,8,2,power_law_1.01,0.14559616088867186
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,8,2,power_law_1.01,0.15758079528808594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,8,2,power_law_1.01,2.1267059326171873
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,8,2,power_law_1.01,0.15676032066345216
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,8,2,power_law_1.01,0.18685951232910156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,8,2,power_law_1.01,0.1914508819580078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,8,2,power_law_1.01,0.1985856056213379
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,8,2,power_law_1.01,0.19923967361450196
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,8,2,power_law_1.01,0.2048409652709961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,8,2,power_law_1.01,0.20773887634277344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,8,2,power_law_1.01,0.21455871582031252
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,8,2,power_law_1.01,0.22615808486938477
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,8,2,power_law_1.01,0.3073510360717774
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,8,2,power_law_1.01,3.1518170166015627
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,8,2,power_law_1.01,0.3590451049804687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,8,2,power_law_1.01,0.4123942565917969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,8,2,power_law_1.01,0.5408332824707032
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,8,2,power_law_1.01,0.6855104064941406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,8,2,power_law_1.01,4.237273559570313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,8,2,power_law_1.01,0.9582272338867188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,8,2,power_law_1.2,0.061607680320739745
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,8,2,power_law_1.01,1.1841356658935547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,8,2,power_law_1.2,0.06773632049560548
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,8,2,power_law_1.2,0.08963968276977538
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,8,2,power_law_1.2,0.10357631683349608
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,8,2,power_law_1.2,0.13451775550842285
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,8,2,power_law_1.2,0.2126848030090332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,8,2,power_law_1.2,0.21806720733642576
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,8,2,power_law_1.2,0.23257984161376952
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,8,2,power_law_1.01,0.148853759765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,8,2,power_law_1.01,1.7597824096679688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,8,2,power_law_1.2,0.2371673583984375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,8,2,power_law_1.2,0.3079948806762695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,8,2,power_law_1.2,0.315098876953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,8,2,power_law_1.2,0.31222015380859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,8,2,power_law_1.2,0.33111038208007815
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,8,2,power_law_1.2,0.3464384078979492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,4,4,power_law_1.2,2.380309753417969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,8,2,power_law_1.2,0.34626686096191406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,8,2,power_law_1.01,2.2499647521972657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,8,2,power_law_1.2,0.3783795166015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,8,2,power_law_1.2,0.42147968292236326
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,8,2,power_law_1.2,0.48453121185302733
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,8,2,power_law_1.2,0.5395558547973633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,8,2,power_law_1.2,0.6096486282348633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,8,2,power_law_1.2,0.7203929901123047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,8,2,power_law_1.2,0.9841689300537109
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,8,2,power_law_1.2,1.2197747039794922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,8,2,power_law_1.2,0.04831744194030761
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,8,2,power_law_1.2,0.0632908821105957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,8,2,power_law_1.2,0.07451007843017579
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,8,2,power_law_1.2,1.6850650024414062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,8,2,power_law_1.2,0.08323328018188476
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,8,2,power_law_1.2,0.14393471717834472
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,8,2,power_law_1.2,0.04106368064880371
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,8,2,power_law_1.2,0.1524441623687744
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,8,2,power_law_1.01,0.25719295501708983
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,8,2,power_law_1.2,0.15601535797119143
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,8,2,power_law_1.2,0.18915712356567382
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,8,2,power_law_1.2,0.19289215087890627
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,8,2,power_law_1.2,0.19890815734863282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,8,2,power_law_1.2,0.20450944900512696
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,8,2,power_law_1.2,0.2110540771484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,8,2,power_law_1.2,2.1927500915527345
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,8,2,power_law_1.2,0.21256832122802732
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,8,2,power_law_1.2,0.21690752029418947
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,8,2,power_law_1.2,0.2263987159729004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,8,2,power_law_1.2,0.27811328887939457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,8,2,power_law_1.2,0.31420032501220707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,8,2,power_law_1.2,0.3694476699829102
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,8,2,power_law_1.2,0.4329792022705078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,8,2,power_law_1.2,0.5669030380249024
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,8,2,power_law_1.2,0.6936345672607421
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,8,2,power_law_1.2,0.15338239669799805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,16,1,balanced,0.06288640022277832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,8,2,power_law_1.2,0.9788953399658202
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,16,1,balanced,0.07400383949279785
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,16,1,balanced,0.13297856330871582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,16,1,balanced,0.2291321563720703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,8,2,power_law_1.2,3.2488641357421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,16,1,balanced,0.30523391723632814
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,16,1,balanced,0.3074841690063477
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,16,1,balanced,0.3078848075866699
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,16,1,balanced,0.31162879943847654
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,8,2,power_law_1.2,1.2062220764160156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,16,1,balanced,0.09155327796936034
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,16,1,balanced,0.31311103820800784
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,16,1,balanced,0.31849344253540035
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,16,1,balanced,0.32494976043701174
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,16,1,balanced,0.35634559631347656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,16,1,balanced,0.3550035095214844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,16,1,balanced,0.36382465362548827
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,16,1,balanced,0.3734431838989258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,16,1,balanced,0.40771839141845706
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,16,1,balanced,0.4462873458862305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,16,1,balanced,0.49678783416748046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,16,1,balanced,0.7149561309814454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,16,1,balanced,0.6936946868896484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,8,2,power_law_1.2,4.30331787109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,16,1,balanced,0.9163948822021485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,16,1,balanced,0.04499263763427734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,16,1,balanced,1.2535916900634765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,16,1,balanced,0.05416704177856445
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,16,1,balanced,0.07357759952545166
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,16,1,balanced,0.10827712059020995
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,16,1,balanced,1.6099801635742188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,16,1,balanced,0.1898784065246582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,16,1,balanced,0.22401792526245118
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,16,1,balanced,0.2304928016662598
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,16,1,balanced,0.2272563171386719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,16,1,balanced,0.228985595703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,16,1,balanced,0.23075584411621092
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,8,2,power_law_1.2,2.2991629028320313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,16,1,balanced,2.1615858459472657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,16,1,balanced,0.23258432388305664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,16,1,balanced,0.23536447525024412
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,16,1,balanced,0.30831104278564453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,8,2,power_law_1.2,1.8160537719726562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,16,1,balanced,0.30863487243652343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,16,1,balanced,0.31832256317138674
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,16,1,balanced,0.32416702270507813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,16,1,balanced,0.3446527862548828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,16,1,balanced,0.3609369659423828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,16,1,balanced,0.3990060806274414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,16,1,balanced,0.4417324829101562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,16,1,balanced,3.085008544921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,16,1,balanced,0.6986669158935547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,16,1,balanced,0.31846271514892577
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,16,1,balanced,0.7754637145996094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,16,1,power_law_1.01,0.07392640113830566
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,16,1,balanced,1.150052490234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,16,1,power_law_1.01,0.12085439682006835
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,16,1,balanced,4.012273254394531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,16,1,balanced,1.478112030029297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,16,1,power_law_1.01,0.24066112518310545
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,16,1,power_law_1.01,0.24981248855590818
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,16,1,power_law_1.01,0.25599487304687496
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,16,1,power_law_1.01,0.2683308792114258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,16,1,balanced,2.0130342102050784
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,16,1,power_law_1.01,0.06281599998474122
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,16,1,power_law_1.01,0.278668155670166
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,16,1,power_law_1.01,0.31799232482910156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,16,1,power_law_1.01,0.09167296409606933
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,16,1,power_law_1.01,0.34138366699218753
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,16,1,power_law_1.01,0.31996927261352537
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,16,1,power_law_1.01,0.3469523239135742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,16,1,power_law_1.01,0.35951038360595705
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,16,1,power_law_1.01,0.3616352081298828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,16,1,balanced,2.7143283081054688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,16,1,power_law_1.01,0.15608127593994142
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,16,1,power_law_1.01,0.41312831878662115
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,16,1,power_law_1.01,0.49228160858154296
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,16,1,power_law_1.01,0.5523046493530274
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,16,1,power_law_1.01,0.5942681503295899
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,16,1,power_law_1.01,0.705560302734375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,16,1,power_law_1.01,0.04225471973419189
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,16,1,power_law_1.01,0.054552321434021
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,16,1,power_law_1.01,0.07261631965637208
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,16,1,power_law_1.01,1.1574527740478515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,16,1,power_law_1.01,0.09793536186218262
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,16,1,power_law_1.01,0.12390080451965332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,16,1,power_law_1.01,0.1907916831970215
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,16,1,power_law_1.01,0.1957472038269043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,16,1,power_law_1.01,0.9256614685058594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,16,1,power_law_1.01,0.20829439163208008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,16,1,power_law_1.01,0.2100627136230469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,16,1,power_law_1.01,2.1477343749999998
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,16,1,power_law_1.01,0.21553983688354492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,16,1,power_law_1.01,0.22201215744018557
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,16,1,power_law_1.01,0.29597631454467777
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,16,1,power_law_1.01,1.6693356323242188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,16,1,power_law_1.01,0.30806335449218747
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,16,1,power_law_1.01,0.30965696334838866
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,16,1,power_law_1.01,0.20178752899169922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,16,1,power_law_1.01,0.3109651184082031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,16,1,power_law_1.01,0.32337600708007813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,16,1,power_law_1.01,0.35544639587402344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,16,1,power_law_1.01,3.1012869262695313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,16,1,power_law_1.01,0.3808768081665039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,16,1,power_law_1.01,0.4666099166870117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,16,1,power_law_1.01,0.5393049621582031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,16,1,power_law_1.01,0.6925580596923828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,16,1,power_law_1.01,0.8473779296874999
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,16,1,power_law_1.2,0.06316415786743164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,16,1,power_law_1.01,4.062251586914063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,16,1,power_law_1.2,0.07303296089172363
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,16,1,power_law_1.01,1.1704307556152345
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,16,1,power_law_1.2,0.11234751701354981
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,16,1,power_law_1.2,0.14779135704040527
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,16,1,power_law_1.2,0.23644351959228516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,16,1,power_law_1.2,0.2600134468078613
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,16,1,power_law_1.2,0.2750540733337402
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,16,1,power_law_1.2,0.09066816329956054
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,16,1,power_law_1.01,2.1165843200683594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,16,1,power_law_1.2,0.3377312088012695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,16,1,power_law_1.01,1.4768704223632814
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,16,1,power_law_1.2,0.33988224029541014
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,16,1,power_law_1.2,0.3506208038330078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,16,1,power_law_1.2,0.37274688720703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,16,1,power_law_1.2,0.391978874206543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,16,1,power_law_1.01,2.7614700317382814
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,16,1,power_law_1.2,0.42705150604248043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,16,1,power_law_1.2,0.24728960037231446
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,16,1,power_law_1.2,0.3503225708007812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,16,1,power_law_1.2,0.5160287857055664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,16,1,power_law_1.2,0.557127685546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,16,1,power_law_1.2,0.29024511337280273
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,16,1,power_law_1.2,0.6167712020874023
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,16,1,power_law_1.2,0.7163283538818359
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,16,1,power_law_1.2,0.04235648155212403
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,16,1,power_law_1.2,0.946078109741211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,16,1,power_law_1.2,0.055591678619384764
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,16,1,power_law_1.2,1.1766783905029297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,16,1,power_law_1.2,0.091211519241333
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,16,1,power_law_1.2,0.11853119850158692
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,16,1,power_law_1.2,0.18647615432739256
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,16,1,power_law_1.2,0.19515840530395506
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,16,1,power_law_1.2,0.2045599937438965
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,16,1,power_law_1.2,0.20980031967163085
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,16,1,power_law_1.2,1.6590757751464842
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,16,1,power_law_1.2,0.2141337585449219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,16,1,power_law_1.2,0.07200895786285401
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,16,1,power_law_1.2,0.22099008560180664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,16,1,power_law_1.2,0.22440959930419924
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,16,1,power_law_1.2,0.3044787216186523
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,16,1,power_law_1.2,0.3065478324890137
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,16,1,power_law_1.2,0.31366464614868167
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,16,1,power_law_1.2,0.32278335571289063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,16,1,power_law_1.2,0.3337971115112305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,16,1,power_law_1.2,0.3593734359741211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,16,1,power_law_1.2,2.145013122558594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,16,1,power_law_1.2,0.3915200042724609
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,16,1,power_law_1.2,0.47828929901123046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,16,1,power_law_1.2,0.5443507385253906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,16,1,power_law_1.2,0.7023052978515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,1,16,balanced,0.06936192035675048
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,16,1,power_law_1.2,0.8526918029785155
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,1,16,balanced,0.06922624111175538
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,1,16,balanced,0.08012288093566895
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,1,16,balanced,0.10893440246582031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,1,16,balanced,0.20448768615722654
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,16,1,power_law_1.2,1.1766342163085937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,1,16,balanced,0.4233407974243164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,1,16,balanced,0.5360588836669922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,1,16,balanced,0.8054412841796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,16,1,power_law_1.2,1.488089599609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,1,16,balanced,0.8072652435302734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,1,16,balanced,0.8074265289306641
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,1,16,balanced,0.8194278717041016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,1,16,balanced,0.8185919952392577
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,1,16,balanced,0.8226573181152343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,1,16,balanced,0.8196160125732422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,1,16,balanced,0.8225638580322265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,1,16,balanced,0.825511703491211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,1,16,balanced,0.84434814453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,1,16,balanced,0.8712371063232421
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,1,16,balanced,0.9026457977294922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,16,1,power_law_1.2,3.114709167480469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,1,16,balanced,0.9494374084472657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,1,16,balanced,1.012284164428711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,1,16,balanced,1.0790220642089845
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,1,16,balanced,1.4194816589355468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,1,16,balanced,0.04295551776885986
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,1,16,balanced,1.4907583618164062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,1,16,balanced,0.04178304195404053
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,1,16,balanced,0.050900478363037106
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,1,16,balanced,2.012450561523438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,1,16,balanced,0.06637311935424804
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,1,16,balanced,0.10816767692565918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,1,16,balanced,0.19460479736328123
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,16,1,power_law_1.2,2.1248902893066406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,1,16,balanced,0.2763737678527832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,1,16,balanced,0.305664005279541
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,1,16,balanced,2.647052917480469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,1,16,balanced,0.30760576248168947
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,1,16,balanced,0.3076915168762207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,1,16,balanced,0.31221248626708986
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,1,16,balanced,0.3158143997192383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,1,16,balanced,0.31765504837036135
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,1,16,balanced,0.31134208679199216
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,1,16,balanced,0.3153100776672363
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,1,16,balanced,0.3148415946960449
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,1,16,balanced,0.3190988731384278
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,1,16,balanced,0.3349375915527344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,1,16,balanced,0.35688831329345705
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,1,16,balanced,3.7737921142578124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,1,16,balanced,0.3937638473510742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,1,16,balanced,0.41576576232910156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,1,16,balanced,0.46854400634765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,1,16,balanced,0.5306240081787109
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,1,16,balanced,0.6685990142822266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,1,16,power_law_1.01,0.20185728073120118
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,1,16,balanced,0.8844083404541015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,1,16,power_law_1.01,0.4215283203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,1,16,balanced,1.1095884704589845
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,1,16,power_law_1.01,0.5182988739013672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,1,16,power_law_1.01,0.4965644836425781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,1,16,balanced,1.4741529846191406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,1,16,power_law_1.01,0.52714111328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,1,16,power_law_1.01,0.8058841705322266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,1,16,power_law_1.01,0.7270259094238282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,1,16,power_law_1.01,0.7720793914794922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,1,16,power_law_1.01,0.7403533172607422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,1,16,power_law_1.01,0.7336383819580078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,1,16,power_law_1.01,0.7123391723632813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,1,16,power_law_1.01,0.7275059509277344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,1,16,power_law_1.01,0.7381043243408203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,1,16,power_law_1.01,0.6633817291259765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,1,16,power_law_1.01,0.7888550567626953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,1,16,power_law_1.01,0.719056625366211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,1,16,power_law_1.01,0.7947148895263672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,1,16,power_law_1.01,0.8295500946044921
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,1,16,power_law_1.01,0.8584076690673829
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,1,16,power_law_1.01,0.9886566162109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,16,1,power_law_1.2,4.084072265625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,1,16,power_law_1.01,1.1619353485107422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,1,16,power_law_1.01,1.520148468017578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,1,16,power_law_1.01,0.10682623863220216
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,1,16,power_law_1.01,1.7392985534667968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,1,16,power_law_1.01,0.19099647521972657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,1,16,power_law_1.01,0.27235456466674807
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,1,16,power_law_1.01,0.2704153633117676
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,1,16,power_law_1.01,0.2753574371337891
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,1,16,power_law_1.01,0.2922265625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,1,16,power_law_1.01,2.539622344970703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,1,16,power_law_1.01,0.2769343948364258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,1,16,power_law_1.01,0.2892006492614746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,1,16,power_law_1.01,0.27123199462890624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,1,16,power_law_1.01,0.28057600021362306
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,1,16,power_law_1.01,0.2971481513977051
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,1,16,power_law_1.01,0.28555648803710937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,1,16,power_law_1.01,0.2929228782653809
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,1,16,power_law_1.01,3.341352844238281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,16,1,power_law_1.2,2.769498291015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,1,16,power_law_1.01,0.252303352355957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,1,16,power_law_1.01,0.29697664260864254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,1,16,power_law_1.01,0.27505792617797853
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,1,16,power_law_1.01,0.29570047378540043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,1,16,power_law_1.01,0.31133056640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,1,16,power_law_1.01,0.3619007873535156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,1,16,power_law_1.01,0.39276031494140623
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,1,16,power_law_1.01,0.44602878570556637
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,1,16,power_law_1.01,4.288875427246094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,1,16,power_law_1.01,0.5610457611083984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,1,16,power_law_1.01,0.6628096008300781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,1,16,power_law_1.2,0.20171648025512695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,1,16,power_law_1.01,0.9351334381103517
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,1,16,power_law_1.2,0.42513278961181644
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,1,16,power_law_1.2,0.4932710266113281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,1,16,power_law_1.01,1.145132827758789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,1,16,power_law_1.2,0.4949964904785157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,1,16,power_law_1.01,5.5194482421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,1,16,power_law_1.2,0.5159257507324219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,1,16,power_law_1.2,0.7463756561279297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,1,16,power_law_1.01,1.7988581848144531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,1,16,power_law_1.2,0.7168985748291016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,1,16,power_law_1.01,2.0692903137207033
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,1,16,power_law_1.2,0.7378892517089843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,1,16,power_law_1.2,0.7168102264404297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,1,16,power_law_1.2,0.7304576110839844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,1,16,power_law_1.2,0.7357798767089844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,1,16,power_law_1.2,0.6979097747802735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,1,16,power_law_1.2,0.77085693359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,1,16,power_law_1.2,0.708753890991211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,1,16,power_law_1.2,0.7118771362304688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,1,16,power_law_1.2,0.7047526550292968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,1,16,power_law_1.2,0.8080665588378906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,1,16,power_law_1.2,0.8117286682128906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,1,16,power_law_1.2,0.9301683044433593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,1,16,power_law_1.2,1.0362150573730469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,1,16,power_law_1.2,1.168737258911133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,1,16,power_law_1.2,0.10666624069213868
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,1,16,power_law_1.2,0.19099391937255858
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,1,16,power_law_1.2,1.6504205322265626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,1,16,power_law_1.2,2.0110003662109373
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,1,16,power_law_1.2,0.25955968856811523
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,1,16,power_law_1.2,0.27699583053588867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,1,16,power_law_1.2,0.2740492820739746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,1,16,power_law_1.2,2.8893695068359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,1,16,power_law_1.2,0.2750873565673828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,1,16,power_law_1.2,0.2764633560180664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,1,16,power_law_1.2,0.2754931259155273
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,1,16,power_law_1.2,0.2826508712768555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,1,16,power_law_1.2,0.2794291114807129
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,1,16,power_law_1.2,0.2954790306091309
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,1,16,power_law_1.2,4.205379943847656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,1,16,power_law_1.2,0.2687027168273926
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,1,16,power_law_1.2,0.25674623489379883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,1,16,power_law_1.2,0.26437631607055667
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,1,16,power_law_1.2,0.2727231979370117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,1,16,power_law_1.2,0.2966720008850098
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,1,16,power_law_1.2,0.326769905090332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,1,16,power_law_1.2,0.37896705627441407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,1,16,power_law_1.2,5.136952514648438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,1,16,power_law_1.2,0.41519615173339847
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,1,16,power_law_1.2,0.4833292770385742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,1,16,power_law_1.2,0.6183654403686523
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,1,16,power_law_1.2,0.29626752853393556
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,2,8,balanced,0.058535680770874024
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,2,8,balanced,0.06649600028991699
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,1,16,power_law_1.2,0.7753318023681641
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,2,8,balanced,0.07812479972839356
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,2,8,balanced,0.11570176124572754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,1,16,power_law_1.2,1.0401203155517578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,2,8,balanced,0.1897702407836914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,2,8,balanced,0.4173977661132812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,2,8,balanced,0.5255104064941406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,1,16,power_law_1.2,1.5087167358398437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,2,8,balanced,0.7853644561767579
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,1,16,power_law_1.2,7.128143310546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,2,8,balanced,0.7875186920166015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,1,16,power_law_1.2,2.135819549560547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,2,8,balanced,0.7864089965820312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,2,8,balanced,0.7987814331054688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,2,8,balanced,0.7972736358642578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,2,8,balanced,0.8085453033447265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,1,16,power_law_1.2,2.6806246948242185
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,2,8,balanced,0.8155609893798829
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,2,8,balanced,0.7804249572753906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,2,8,balanced,0.84123779296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,2,8,balanced,0.8764288330078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,2,8,balanced,0.9134400177001953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,2,8,balanced,0.9947212982177736
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,2,8,balanced,1.0823577880859374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,2,8,balanced,1.2047334289550782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,2,8,balanced,0.03481215953826904
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,2,8,balanced,0.0432038402557373
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,2,8,balanced,1.4257728576660156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,2,8,balanced,0.05118336200714111
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,2,8,balanced,0.06967679977416992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,2,8,balanced,0.11331839561462402
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,2,8,balanced,1.7237107849121096
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,2,8,balanced,0.19689855575561524
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,2,8,balanced,0.2824051284790039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,2,8,balanced,0.30679168701171877
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,2,8,balanced,0.3081177520751953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,2,8,balanced,0.3095667266845703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,2,8,balanced,0.8246284484863281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,2,8,balanced,0.30281087875366214
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,2,8,balanced,2.2937164306640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,2,8,balanced,0.30727680206298824
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,2,8,balanced,0.3098534393310547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,2,8,balanced,0.31582208633422854
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,2,8,balanced,0.32466175079345705
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,2,8,balanced,0.33090431213378907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,2,8,balanced,0.35368961334228516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,2,8,balanced,0.3748556900024414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,2,8,balanced,0.40089599609375004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,2,8,balanced,0.44879104614257814
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,2,8,balanced,0.49293056488037107
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,2,8,balanced,3.115914306640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,2,8,balanced,0.5935744094848633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,2,8,balanced,0.7028813171386719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,2,8,power_law_1.01,0.11920639991760254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,2,8,power_law_1.01,0.1846182441711426
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,2,8,balanced,0.8942054748535156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,2,8,power_law_1.01,0.38254974365234373
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,2,8,power_law_1.01,0.40814334869384766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,2,8,power_law_1.01,0.44683135986328126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,2,8,power_law_1.01,0.46992641448974604
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,2,8,balanced,1.5997503662109376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,2,8,power_law_1.01,0.4992524719238281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,2,8,power_law_1.01,0.7817024230957031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,2,8,balanced,2.1280972290039064
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,2,8,balanced,1.208684768676758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,2,8,power_law_1.01,0.7732339477539062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,2,8,power_law_1.01,0.7841689300537109
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,2,8,power_law_1.01,0.706719970703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,2,8,balanced,4.324300842285156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,2,8,power_law_1.01,0.7233420562744141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,2,8,power_law_1.01,0.7308812713623046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,2,8,power_law_1.01,0.7183564758300781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,2,8,power_law_1.01,0.756412124633789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,2,8,power_law_1.01,0.7532518768310548
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,2,8,power_law_1.01,0.8001920318603515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,2,8,power_law_1.01,0.8282189178466798
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,2,8,power_law_1.01,0.8750924682617187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,2,8,power_law_1.01,0.9874982452392578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,2,8,power_law_1.01,1.0758477020263673
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,2,8,power_law_1.01,0.06744192123413087
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,2,8,power_law_1.01,1.4124928283691407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,2,8,power_law_1.01,0.10993023872375489
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,2,8,power_law_1.01,1.6185061645507812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,2,8,power_law_1.01,0.19254911422729493
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,2,8,power_law_1.01,2.0308749389648435
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,2,8,power_law_1.01,0.16618879318237306
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,2,8,power_law_1.01,0.2283660888671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,2,8,power_law_1.01,0.24811391830444335
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,2,8,power_law_1.01,0.26798847198486325
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,2,8,power_law_1.01,0.28609664916992184
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,2,8,power_law_1.01,0.2689625549316406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,2,8,power_law_1.01,2.735274353027344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,2,8,power_law_1.01,0.2738073539733887
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,2,8,power_law_1.01,0.28109695434570314
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,2,8,power_law_1.01,0.27673599243164065
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,2,8,power_law_1.01,0.278656005859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,2,8,power_law_1.01,0.28000511169433595
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,2,8,power_law_1.01,0.28936960220336916
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,2,8,power_law_1.01,0.31264896392822267
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,2,8,power_law_1.01,0.32780799865722654
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,2,8,power_law_1.01,0.35841022491455077
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,2,8,power_law_1.01,3.7683520507812505
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,2,8,power_law_1.01,0.4230694580078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,2,8,power_law_1.01,0.4918809509277343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,2,8,power_law_1.01,0.2708620834350586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,2,8,power_law_1.01,0.5964748764038086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,2,8,power_law_1.01,0.6803801727294922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,2,8,power_law_1.2,0.11783040046691895
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,2,8,power_law_1.01,4.804007568359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,2,8,power_law_1.2,0.18618751525878907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,2,8,power_law_1.01,1.1575385284423827
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,2,8,power_law_1.2,0.3309734344482422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,2,8,power_law_1.2,0.4451942443847656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,2,8,power_law_1.01,1.6264012145996094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,2,8,power_law_1.2,0.476416015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,2,8,power_law_1.2,0.46722942352294916
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,2,8,power_law_1.01,0.930535659790039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,2,8,power_law_1.2,0.7807987213134766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,2,8,power_law_1.2,0.30277887344360355
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,2,8,power_law_1.2,0.5999795150756836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,2,8,power_law_1.01,2.224648895263672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,2,8,power_law_1.2,0.7131724548339844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,2,8,power_law_1.2,0.7812646484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,2,8,power_law_1.2,0.7250163269042968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,2,8,power_law_1.2,0.7114444732666015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,2,8,power_law_1.2,0.705818862915039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,2,8,power_law_1.2,0.7418291473388672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,2,8,power_law_1.2,0.7671014404296874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,2,8,power_law_1.2,0.8573862457275391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,2,8,power_law_1.2,0.8334719848632812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,2,8,power_law_1.2,0.9074124908447265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,2,8,power_law_1.2,0.9840115356445314
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,2,8,power_law_1.2,1.130108184814453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,2,8,power_law_1.2,0.06788352012634277
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,2,8,power_law_1.2,1.4958963012695312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,2,8,power_law_1.2,0.1102937602996826
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,2,8,power_law_1.2,0.15768447875976563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,2,8,power_law_1.2,0.17349119186401368
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,2,8,power_law_1.2,1.855731201171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,2,8,power_law_1.2,0.20552192687988283
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,2,8,power_law_1.2,0.2556480026245117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,2,8,power_law_1.2,2.239471435546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,2,8,power_law_1.2,0.25135488510131837
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,2,8,power_law_1.2,0.30623872756958004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,2,8,power_law_1.2,0.24692607879638673
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,2,8,power_law_1.2,0.28451072692871093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,2,8,power_law_1.2,0.27679487228393557
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,2,8,power_law_1.2,0.27149951934814454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,2,8,power_law_1.2,2.9253118896484374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,2,8,power_law_1.2,0.27777919769287107
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,2,8,power_law_1.2,0.2855014419555664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,2,8,power_law_1.2,0.2918016052246094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,2,8,power_law_1.2,0.3217356872558594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,2,8,power_law_1.2,0.35618431091308594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,2,8,power_law_1.2,0.3634739303588867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,2,8,power_law_1.2,0.4472281646728516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,2,8,power_law_1.2,0.5145267105102539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,2,8,power_law_1.2,4.301085510253906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,2,8,power_law_1.2,0.6197951889038086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,2,8,power_law_1.2,0.7818214416503906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,4,4,balanced,0.0541644811630249
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,2,8,power_law_1.2,0.26092159271240234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,4,4,balanced,0.07951360225677491
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,4,4,balanced,0.06363776206970215
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,4,4,balanced,0.11003647804260255
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,4,4,balanced,0.18720384597778322
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,4,4,balanced,0.4078348922729492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,2,8,power_law_1.2,1.2852146911621094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,4,4,balanced,0.5170483016967774
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,2,8,power_law_1.2,5.855955200195313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,4,4,balanced,0.5210124969482421
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,4,4,balanced,0.5239769744873046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,2,8,power_law_1.2,1.031031036376953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,2,8,power_law_1.2,1.9349644470214844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,4,4,balanced,0.7700415802001953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,4,4,balanced,0.78761474609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,4,4,balanced,0.7792435455322265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,2,8,power_law_1.2,2.4236428833007815
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,4,4,balanced,0.7948722839355469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,4,4,balanced,0.8028467559814454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,4,4,balanced,0.5267647933959961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,4,4,balanced,0.8144358062744141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,4,4,balanced,0.8312640380859374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,4,4,balanced,0.8570905303955078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,4,4,balanced,0.9029094696044921
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,4,4,balanced,0.9743296051025391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,4,4,balanced,1.0557119750976562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,4,4,balanced,0.03426687955856324
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,4,4,balanced,1.1867276763916015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,4,4,balanced,0.03972095966339111
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,4,4,balanced,0.05125631809234619
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,4,4,balanced,0.07264895915985108
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,4,4,balanced,0.1174899196624756
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,4,4,balanced,1.734862060546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,4,4,balanced,0.20280319213867187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,4,4,balanced,0.29380992889404295
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,4,4,balanced,0.29556095123291015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,4,4,balanced,0.2982694435119629
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,4,4,balanced,0.29851903915405276
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,4,4,balanced,2.2674688720703124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,4,4,balanced,1.4462655639648436
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,4,4,balanced,0.33440513610839845
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,4,4,balanced,0.33667072296142575
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,4,4,balanced,0.34051071166992186
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,4,4,balanced,0.3457036972045898
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,4,4,balanced,0.3614336013793945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,4,4,balanced,0.35994110107421873
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,4,4,balanced,0.3721100616455078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,4,4,balanced,0.39710720062255855
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,4,4,balanced,0.41924480438232425
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,4,4,balanced,0.4629900741577148
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,4,4,balanced,3.072157287597656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,4,4,balanced,0.5079078292846679
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,4,4,balanced,0.600181770324707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,4,4,power_law_1.01,0.07988736152648926
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,4,4,balanced,0.6933503723144531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,4,4,balanced,0.8835545349121094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,4,4,power_law_1.01,0.10883328437805176
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,4,4,power_law_1.01,0.17734527587890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,4,4,power_law_1.01,0.19194112777709962
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,4,4,balanced,1.2058636474609377
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,4,4,balanced,4.245342712402343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,4,4,power_law_1.01,0.28114559173583986
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,4,4,power_law_1.01,0.40030464172363284
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,4,4,power_law_1.01,0.4189260864257813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,4,4,balanced,1.5730995178222655
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,4,4,power_law_1.01,0.48411647796630863
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,4,4,power_law_1.01,0.45968383789062506
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,4,4,power_law_1.01,0.48623489379882817
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,4,4,balanced,2.105032958984375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,4,4,power_law_1.01,0.6976767730712891
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,4,4,power_law_1.01,0.6808217620849609
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,4,4,power_law_1.01,0.6932351684570313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,4,4,power_law_1.01,0.7190579223632813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,4,4,power_law_1.01,0.780684814453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,4,4,power_law_1.01,0.7657049560546876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,4,4,power_law_1.01,0.8037427520751953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,4,4,power_law_1.01,0.854940185546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,4,4,power_law_1.01,0.6935167694091797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,4,4,power_law_1.01,0.9241024017333984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,4,4,power_law_1.01,1.0713152313232421
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,4,4,power_law_1.01,1.2558451080322266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,4,4,power_law_1.01,0.07041024208068848
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,4,4,power_law_1.01,1.4670848083496093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,4,4,power_law_1.01,0.11343487739562988
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,4,4,power_law_1.01,0.13034239768981934
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,4,4,power_law_1.01,0.16941823959350585
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,4,4,power_law_1.01,1.9796723937988283
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,4,4,power_law_1.01,0.05076735973358154
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,4,4,power_law_1.01,0.22928255081176757
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,4,4,power_law_1.01,0.2630489540100098
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,4,4,power_law_1.01,0.28724351882934573
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,4,4,power_law_1.01,0.2689945602416992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,4,4,power_law_1.01,2.3988467407226564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,4,4,power_law_1.01,0.3089484786987305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,4,4,power_law_1.01,0.2998003196716309
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,4,4,power_law_1.01,0.3012441635131836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,4,4,power_law_1.01,0.3097049522399902
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,4,4,power_law_1.01,0.3210355377197266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,4,4,power_law_1.01,0.3338790512084961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,4,4,power_law_1.01,0.3584025573730469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,4,4,power_law_1.01,3.5707800292968748
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,4,4,power_law_1.01,0.437256965637207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,4,4,power_law_1.01,0.5017279815673829
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,4,4,power_law_1.01,0.3020454406738281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,4,4,power_law_1.01,0.19324031829833985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,4,4,power_law_1.01,0.5878937530517578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,4,4,power_law_1.01,0.7454182434082031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,4,4,power_law_1.01,4.597397766113281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,4,4,power_law_1.2,0.08000384330749512
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,4,4,power_law_1.01,0.37794559478759765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,4,4,power_law_1.01,0.9662413024902344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,4,4,power_law_1.2,0.15668224334716796
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,4,4,power_law_1.2,0.18005760192871095
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,4,4,power_law_1.2,0.2523980712890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,4,4,power_law_1.01,1.2099788665771485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,4,4,power_law_1.2,0.37878143310546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,4,4,power_law_1.2,0.4165363311767578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,4,4,power_law_1.01,1.7294540405273438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,4,4,power_law_1.2,0.4750643157958985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,4,4,power_law_1.2,0.10845824241638184
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,4,4,power_law_1.2,0.4769126510620117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,4,4,power_law_1.2,0.508108787536621
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,4,4,power_law_1.01,2.131636505126953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,4,4,power_law_1.2,0.699705581665039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,4,4,power_law_1.2,0.6885298919677735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,4,4,power_law_1.2,0.6842752075195312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,4,4,power_law_1.2,0.7098047637939453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,4,4,power_law_1.2,0.7364466857910157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,4,4,power_law_1.2,0.760843505859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,4,4,power_law_1.2,0.7691187286376954
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,4,4,power_law_1.2,0.8188185882568358
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,4,4,power_law_1.2,0.8722841644287109
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,4,4,power_law_1.2,0.9710681915283204
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,4,4,power_law_1.2,1.1078195190429687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,4,4,power_law_1.2,0.05066495895385742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,4,4,power_law_1.2,1.318219451904297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,4,4,power_law_1.2,0.0699724817276001
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,4,4,power_law_1.2,0.09817343711853027
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,4,4,power_law_1.2,1.5513523864746095
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,4,4,power_law_1.2,0.12118144035339355
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,4,4,power_law_1.2,0.15710335731506347
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,4,4,power_law_1.2,0.1881932830810547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,4,4,power_law_1.2,1.979479064941406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,4,4,power_law_1.2,0.21886720657348632
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,4,4,power_law_1.2,0.2638809585571289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,4,4,power_law_1.2,0.27016191482543944
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,4,4,power_law_1.2,0.2586777687072754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,4,4,power_law_1.2,0.30586368560791016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,4,4,power_law_1.2,2.61828857421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,4,4,power_law_1.2,0.29666559219360356
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,4,4,power_law_1.2,0.3024563217163086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,4,4,power_law_1.2,0.3089254379272461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,4,4,power_law_1.2,0.3166092872619629
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,4,4,power_law_1.2,0.3323775863647461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,4,4,power_law_1.2,0.33379070281982426
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,4,4,power_law_1.2,0.3634585571289063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,4,4,power_law_1.2,0.3897715377807617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,4,4,power_law_1.2,0.4676544189453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,4,4,power_law_1.2,3.616611938476562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,4,4,power_law_1.2,0.5280652618408203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,8,2,balanced,0.056266241073608406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,4,4,power_law_1.2,0.6497305297851563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,4,4,power_law_1.2,0.7882355499267578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,8,2,balanced,0.06428671836853027
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,8,2,balanced,0.07674880027770996
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,4,4,power_law_1.2,0.9843020629882812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,8,2,balanced,0.1112396812438965
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,8,2,balanced,0.1892313575744629
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,4,4,power_law_1.2,1.2581145477294922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,8,2,balanced,0.41467777252197263
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,4,4,power_law_1.2,4.760641174316406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,8,2,balanced,0.5185599899291992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,8,2,balanced,0.5241996765136718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,8,2,balanced,0.5257344055175781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,8,2,balanced,0.528084487915039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,8,2,balanced,0.5341017532348633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,4,4,power_law_1.2,1.8340313720703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,8,2,balanced,0.5457164764404296
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,8,2,balanced,0.5405337524414062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,8,2,balanced,0.7767922973632813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,8,2,balanced,0.7835622406005859
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,4,4,power_law_1.2,2.3129458618164063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,8,2,balanced,0.7876044464111328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,8,2,balanced,0.8070259094238281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,8,2,balanced,0.8386201477050781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,8,2,balanced,0.8748966217041015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,8,2,balanced,0.9337766265869141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,8,2,balanced,0.033058559894561766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,8,2,balanced,1.0235456085205077
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,8,2,balanced,1.1833420562744141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,8,2,balanced,0.040887041091918944
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,8,2,balanced,0.05532159805297852
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,8,2,balanced,0.07781760215759277
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,8,2,balanced,0.13411199569702148
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,8,2,balanced,1.5217485046386718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,8,2,balanced,0.24329727172851562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,8,2,balanced,0.343059196472168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,8,2,balanced,1.8301528930664062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,8,2,balanced,0.34898048400878906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,8,2,balanced,0.3503590393066406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,8,2,balanced,0.3523891067504883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,8,2,balanced,0.35786880493164064
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,8,2,balanced,0.36108543395996096
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,8,2,balanced,0.42703617095947266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,8,2,balanced,0.433196792602539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,8,2,balanced,0.43814655303955075
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,8,2,balanced,0.44756607055664066
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,8,2,balanced,0.3469440078735352
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,8,2,balanced,0.4623372650146484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,8,2,balanced,3.1257061767578125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,8,2,balanced,0.48290176391601564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,8,2,balanced,2.332757720947266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,8,2,balanced,0.5245849609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,8,2,balanced,0.5679411315917969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,8,2,balanced,0.6537625885009766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,8,2,balanced,0.7410662078857422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,8,2,power_law_1.01,0.06435071945190429
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,8,2,balanced,0.9243238067626953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,8,2,balanced,4.261997985839844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,8,2,power_law_1.01,0.07681280136108398
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,8,2,power_law_1.01,0.10696448326110838
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,8,2,power_law_1.01,0.13201919555664063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,8,2,power_law_1.01,0.2146803283691406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,8,2,balanced,1.3187391662597656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,8,2,power_law_1.01,0.339683837890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,8,2,power_law_1.01,0.3911155319213867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,8,2,balanced,1.6662130737304686
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,8,2,power_law_1.01,0.43861759185791016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,8,2,power_law_1.01,0.4526054382324219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,8,2,power_law_1.01,0.4443353652954102
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,8,2,power_law_1.01,0.46391551971435546
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,8,2,power_law_1.01,0.4663961410522461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,8,2,power_law_1.01,0.47637374877929684
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,8,2,power_law_1.01,0.7037324523925781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,8,2,power_law_1.01,0.713210906982422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,8,2,power_law_1.01,0.7135193634033203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,8,2,power_law_1.01,0.7378265380859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,8,2,power_law_1.01,0.7920665740966797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,8,2,power_law_1.01,0.8284774780273437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,8,2,balanced,2.2363314819335938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,8,2,power_law_1.01,0.9262258911132812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,8,2,power_law_1.01,1.0564134216308594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,8,2,power_law_1.01,1.2524543762207032
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,8,2,power_law_1.01,0.054711041450500486
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,8,2,power_law_1.01,1.4928282165527345
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,8,2,power_law_1.01,0.0764185619354248
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,8,2,power_law_1.01,0.09590527534484863
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,8,2,power_law_1.01,0.1599142360687256
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,8,2,power_law_1.01,0.041880321502685544
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,8,2,power_law_1.01,0.1916851234436035
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,8,2,power_law_1.01,0.214147834777832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,8,2,power_law_1.01,2.3834010314941407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,8,2,power_law_1.01,0.27115392684936523
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,8,2,power_law_1.01,0.27735807418823244
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,8,2,power_law_1.01,0.2812339210510254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,8,2,power_law_1.01,0.29091968536376955
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,8,2,power_law_1.01,0.30542335510253904
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,8,2,power_law_1.01,0.3083955192565918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,8,2,power_law_1.01,1.93662841796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,8,2,power_law_1.01,0.3724160003662109
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,8,2,power_law_1.01,0.3732966232299805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,8,2,power_law_1.01,0.3824076843261719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,8,2,power_law_1.01,3.327736206054688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,8,2,power_law_1.01,0.39038719177246095
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,8,2,power_law_1.01,0.42632190704345707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,8,2,power_law_1.01,0.44851711273193356
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,8,2,power_law_1.01,0.5162188720703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,8,2,power_law_1.01,0.5940723037719726
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,8,2,power_law_1.01,0.6945855712890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,8,2,power_law_1.2,0.06418560028076173
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,8,2,power_law_1.01,4.2725222778320315
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,8,2,power_law_1.2,0.07654016017913819
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,8,2,power_law_1.01,1.102894058227539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,8,2,power_law_1.2,0.10105600357055664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,8,2,power_law_1.2,0.1334822368621826
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,8,2,power_law_1.2,0.2070591926574707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,8,2,power_law_1.01,1.3520716857910156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,8,2,power_law_1.2,0.2545574378967285
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,8,2,power_law_1.2,0.36897537231445315
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,8,2,power_law_1.01,0.8417433929443359
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,8,2,power_law_1.2,0.43432449340820317
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,8,2,power_law_1.01,1.8494528198242186
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,8,2,power_law_1.2,0.4440102386474609
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,8,2,power_law_1.2,0.4313740921020508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,8,2,power_law_1.2,0.4625651168823242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,8,2,power_law_1.2,0.46564990997314454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,8,2,power_law_1.2,0.4716198348999024
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,8,2,power_law_1.01,2.3688563537597656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,8,2,power_law_1.2,0.70063232421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,8,2,power_law_1.2,0.6959129333496094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,8,2,power_law_1.2,0.735999984741211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,8,2,power_law_1.2,0.7932659149169922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,8,2,power_law_1.2,0.8434239959716796
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,8,2,power_law_1.2,0.9406041717529297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,8,2,power_law_1.2,1.0912729644775392
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,8,2,power_law_1.2,0.720232925415039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,8,2,power_law_1.2,1.295098876953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,8,2,power_law_1.2,0.04195712089538574
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,8,2,power_law_1.2,0.05447679996490479
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,8,2,power_law_1.2,0.07270016193389892
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,8,2,power_law_1.2,0.09617664337158202
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,8,2,power_law_1.2,0.13568511962890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,8,2,power_law_1.2,1.9747059631347657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,8,2,power_law_1.2,0.18331520080566405
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,8,2,power_law_1.2,0.2035942459106445
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,8,2,power_law_1.2,0.2626137542724609
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,8,2,power_law_1.2,2.4940185546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,8,2,power_law_1.2,0.28798336029052735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,8,2,power_law_1.2,0.2646963119506836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,8,2,power_law_1.2,0.28019840240478516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,8,2,power_law_1.2,0.30035839080810545
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,8,2,power_law_1.2,0.3089907264709473
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,8,2,power_law_1.2,0.3727667236328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,8,2,power_law_1.2,0.38363391876220704
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,8,2,power_law_1.2,3.4355584716796876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,8,2,power_law_1.2,0.39472896575927735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,8,2,power_law_1.2,1.510734100341797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,8,2,power_law_1.2,0.43328769683837887
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,8,2,power_law_1.2,0.45118976593017573
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,8,2,power_law_1.2,0.5297856140136719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,8,2,power_law_1.2,0.37126144409179684
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,8,2,power_law_1.2,0.6024204635620117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,8,2,power_law_1.2,4.3616180419921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,8,2,power_law_1.2,0.7307775878906251
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,8,2,power_law_1.2,0.8563200378417969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,16,1,balanced,0.0700057601928711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,16,1,balanced,0.07664896011352539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,16,1,balanced,0.09146688461303712
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,8,2,power_law_1.2,1.1197811126708985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,16,1,balanced,0.12772671699523924
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,16,1,balanced,0.20658048629760742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,8,2,power_law_1.2,1.3900915527343751
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,16,1,balanced,0.5699078369140624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,16,1,balanced,0.5740857696533203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,16,1,balanced,0.5747846221923828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,8,2,power_law_1.2,1.9033920288085937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,16,1,balanced,0.5795027160644531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,16,1,balanced,0.586264305114746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,16,1,balanced,0.5946054458618164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,16,1,balanced,0.5996819305419921
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,8,2,power_law_1.2,2.3808396911621097
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,16,1,balanced,0.44924224853515626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,16,1,balanced,0.623001594543457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,16,1,balanced,0.6338310241699219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,16,1,balanced,0.8404038238525391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,16,1,balanced,0.8717485046386718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,16,1,balanced,0.9060723114013672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,16,1,balanced,0.6114559936523437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,16,1,balanced,0.9810009765625001
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,16,1,balanced,1.0531597137451172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,16,1,balanced,0.04562880039215088
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,16,1,balanced,0.058037757873535156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,16,1,balanced,1.2071692657470703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,16,1,balanced,0.07525824069976807
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,16,1,balanced,0.10770688056945801
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,16,1,balanced,0.18049280166625975
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,16,1,balanced,0.3233286285400391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,16,1,balanced,1.8171589660644532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,16,1,balanced,0.46817985534667966
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,16,1,balanced,1.9169229125976563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,16,1,balanced,0.4700032043457031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,16,1,balanced,0.4717132949829102
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,16,1,balanced,0.4763328170776367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,16,1,balanced,0.48266689300537113
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,16,1,balanced,0.4667180633544922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,16,1,balanced,0.48386302947998044
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,16,1,balanced,2.514323272705078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,16,1,balanced,0.4896870422363281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,16,1,balanced,0.4978425598144531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,16,1,balanced,0.5039839935302735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,16,1,balanced,0.6196051025390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,16,1,balanced,0.6396582412719727
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,16,1,balanced,3.2516589355468755
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,16,1,balanced,0.6937670135498047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,16,1,balanced,0.7351417541503906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,16,1,balanced,0.8355865478515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,16,1,balanced,0.9081568145751954
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,16,1,power_law_1.01,0.06789184093475342
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,16,1,balanced,0.6635584259033204
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,16,1,balanced,1.0836844635009766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,16,1,power_law_1.01,0.07600255966186523
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,16,1,power_law_1.01,0.09305983543395996
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,16,1,balanced,4.784579772949219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,16,1,power_law_1.01,0.19234560012817384
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,16,1,balanced,1.647034912109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,16,1,power_law_1.01,0.28404352188110354
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,16,1,power_law_1.01,0.33836734771728516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,16,1,balanced,2.000995788574219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,16,1,power_law_1.01,0.4676671981811523
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,16,1,power_law_1.01,0.4759225463867187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,16,1,power_law_1.01,0.12642687797546387
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,16,1,power_law_1.01,0.4867264175415039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,16,1,power_law_1.01,0.49444606781005856
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,16,1,power_law_1.01,0.5050425720214844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,16,1,power_law_1.01,0.5145196914672852
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,16,1,power_law_1.01,0.5345119857788087
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,16,1,balanced,2.750302734375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,16,1,power_law_1.01,0.5533407974243164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,16,1,power_law_1.01,0.6020256042480469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,16,1,power_law_1.01,0.7684210968017579
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,16,1,power_law_1.01,0.8066617584228515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,16,1,power_law_1.01,0.8555436706542968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,16,1,power_law_1.01,0.9531827545166015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,16,1,power_law_1.01,1.0787564849853515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,16,1,power_law_1.01,0.0454636812210083
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,16,1,power_law_1.01,1.3002291870117186
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,16,1,power_law_1.01,0.05847231864929199
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,16,1,power_law_1.01,1.564281005859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,16,1,power_law_1.01,0.07484543800354004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,16,1,power_law_1.01,0.10796799659729003
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,16,1,power_law_1.01,0.16937280654907227
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,16,1,power_law_1.01,2.0631289672851563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,16,1,power_law_1.01,0.23970111846923828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,16,1,power_law_1.01,0.2712524795532226
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,16,1,power_law_1.01,0.3320595169067383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,16,1,power_law_1.01,2.5191334533691405
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,16,1,power_law_1.01,0.35322879791259765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,16,1,power_law_1.01,0.37412353515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,16,1,power_law_1.01,0.3835961532592774
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,16,1,power_law_1.01,0.390463981628418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,16,1,power_law_1.01,0.40306430816650385
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,16,1,power_law_1.01,0.4140755081176758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,16,1,power_law_1.01,0.43253246307373044
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,16,1,power_law_1.01,3.442742919921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,16,1,power_law_1.01,0.5384115219116211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,16,1,power_law_1.01,0.5691641616821289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,16,1,power_law_1.01,0.5996921539306641
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,16,1,power_law_1.01,0.4201996612548828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,16,1,power_law_1.01,0.7462783813476562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,16,1,power_law_1.01,4.412101135253907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,16,1,power_law_1.01,0.8881734466552734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,16,1,power_law_1.2,0.06754816055297852
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,16,1,power_law_1.01,0.6644850921630859
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,16,1,power_law_1.2,0.07646975994110108
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,16,1,power_law_1.01,1.0262118530273436
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,16,1,power_law_1.2,0.09155263900756835
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,16,1,power_law_1.2,0.12568639755249023
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,16,1,power_law_1.2,0.18099647521972656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,16,1,power_law_1.2,0.26560895919799804
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,16,1,power_law_1.01,1.6177055358886718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,16,1,power_law_1.2,0.31428543090820316
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,16,1,power_law_1.2,0.4635315322875977
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,16,1,power_law_1.01,1.3364300537109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,16,1,power_law_1.2,0.4733734512329102
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,16,1,power_law_1.01,2.232845458984375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,16,1,power_law_1.2,0.48925697326660156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,16,1,power_law_1.2,0.5033343887329101
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,16,1,power_law_1.2,0.5102342224121094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,16,1,power_law_1.2,0.5296588897705078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,16,1,power_law_1.2,0.5567206573486329
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,16,1,power_law_1.2,0.6014508819580078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,16,1,power_law_1.01,2.851529541015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,16,1,power_law_1.2,0.4785043334960937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,16,1,power_law_1.2,0.7677574157714844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,16,1,power_law_1.2,0.8166035461425782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,16,1,power_law_1.2,0.8629043579101563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,16,1,power_law_1.2,0.9814252471923828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,16,1,power_law_1.2,1.1198143768310547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,16,1,power_law_1.2,0.0448851203918457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,16,1,power_law_1.2,1.3297830200195313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,16,1,power_law_1.2,0.0580838394165039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,16,1,power_law_1.2,0.07582143783569337
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,16,1,power_law_1.2,1.608008270263672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,16,1,power_law_1.2,0.10785792350769044
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,16,1,power_law_1.2,0.15987839698791503
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,16,1,power_law_1.2,2.057752990722656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,16,1,power_law_1.2,0.2342745590209961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,16,1,power_law_1.2,0.25875648498535153
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,16,1,power_law_1.2,0.32687103271484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,16,1,power_law_1.2,2.5531015014648437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,16,1,power_law_1.2,0.3407244873046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,16,1,power_law_1.2,0.35825855255126954
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,16,1,power_law_1.2,0.371868782043457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,16,1,power_law_1.2,0.38900222778320315
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,16,1,power_law_1.2,0.3979072189331055
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,16,1,power_law_1.2,0.4115296173095703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,16,1,power_law_1.2,0.42055103302001956
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,16,1,power_law_1.2,0.43118144989013674
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,16,1,power_law_1.2,0.5424953460693359
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,16,1,power_law_1.2,0.579920654296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,16,1,power_law_1.2,0.6105625534057617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,16,1,power_law_1.2,0.6943110656738282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,16,1,power_law_1.2,0.7835622406005859
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,16,1,power_law_1.2,0.9274208068847656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,16,1,power_law_1.2,3.4746270751953126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,1,16,balanced,0.025377280712127685
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,1,16,balanced,0.02392319917678833
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,16,1,power_law_1.2,1.0883660888671876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,16,1,power_law_1.2,4.419728088378906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,1,16,balanced,0.042051839828491214
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,1,16,balanced,0.050553598403930665
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,1,16,balanced,0.0747711992263794
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,1,16,balanced,0.13077247619628907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,1,16,balanced,0.16661247253417968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,1,16,balanced,0.17124095916748047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,16,1,power_law_1.2,1.3879685974121094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,1,16,balanced,0.17117439270019533
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,1,16,balanced,0.1775129508972168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,1,16,balanced,0.1813030433654785
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,1,16,balanced,0.18494464874267577
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,1,16,balanced,0.18861696243286133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,1,16,balanced,0.1861926460266113
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,1,16,balanced,0.19305599212646485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,1,16,balanced,0.2087014389038086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,16,1,power_law_1.2,1.688118438720703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,1,16,balanced,0.2257574462890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,1,16,balanced,0.25726463317871096
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,1,16,balanced,0.28166784286499025
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,1,16,balanced,0.1666022491455078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,1,16,balanced,0.3373516845703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,1,16,balanced,0.37838207244873046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,1,16,balanced,0.5325696182250976
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,1,16,balanced,0.18480255126953124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,16,1,power_law_1.2,2.2974624633789062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,1,16,balanced,0.6625011444091797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,1,16,balanced,0.025708799362182615
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,1,16,balanced,0.025830399990081788
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,1,16,balanced,0.027443199157714843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,1,16,balanced,0.036048638820648196
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,1,16,balanced,0.962914581298828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,1,16,balanced,0.08326272010803223
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,1,16,balanced,0.0928934383392334
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,1,16,balanced,0.09261183738708496
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,1,16,balanced,0.09651583671569824
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,1,16,balanced,0.09696255683898926
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,1,16,balanced,1.2330022430419922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,1,16,balanced,0.099749116897583
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,1,16,balanced,0.1043494415283203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,1,16,balanced,0.09632767677307129
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,1,16,balanced,0.10142592430114745
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,16,1,power_law_1.2,2.9253900146484377
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,1,16,balanced,0.10527999877929688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,1,16,balanced,0.11118720054626466
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,1,16,balanced,0.12836480140686035
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,1,16,balanced,0.1461952018737793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,1,16,balanced,0.19338367462158204
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,1,16,balanced,0.050348801612854
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,1,16,balanced,0.2275071907043457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,1,16,balanced,0.31693952560424804
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,1,16,balanced,0.383164176940918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,1,16,balanced,0.10152447700500489
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,1,16,balanced,0.5492134475708008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,1,16,balanced,0.10003328323364258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,1,16,balanced,0.7057804870605469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,1,16,power_law_1.01,0.12176383972167967
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,1,16,power_law_1.01,0.11908479690551758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,1,16,power_law_1.01,0.12344448089599609
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,1,16,balanced,0.16007936477661133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,1,16,power_law_1.01,0.11125375747680663
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,1,16,power_law_1.01,0.12667648315429686
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,1,16,power_law_1.01,0.13213695526123045
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,1,16,power_law_1.01,0.1364915180206299
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,1,16,power_law_1.01,0.13718015670776368
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,1,16,power_law_1.01,0.1645248031616211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,1,16,power_law_1.01,0.16030336380004884
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,1,16,power_law_1.01,0.16115968704223632
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,1,16,power_law_1.01,0.1622118377685547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,1,16,power_law_1.01,0.20642047882080078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,1,16,power_law_1.01,0.07037055969238282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,1,16,power_law_1.01,0.26425600051879883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,1,16,power_law_1.01,0.29801727294921876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,1,16,power_law_1.01,0.33107967376708985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,1,16,power_law_1.01,0.3798720169067383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,1,16,power_law_1.01,0.1246604824066162
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,1,16,power_law_1.01,0.5014886474609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,1,16,power_law_1.01,0.6848627471923828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,1,16,power_law_1.01,0.18144512176513672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,1,16,power_law_1.01,0.8780480194091798
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,1,16,power_law_1.01,0.07987840175628662
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,1,16,power_law_1.01,0.08021247863769532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,1,16,power_law_1.01,0.08152704238891602
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,1,16,power_law_1.01,0.08165120124816895
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,1,16,power_law_1.01,1.2927935791015623
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,1,16,power_law_1.01,0.06292863845825195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,1,16,power_law_1.01,0.08918911933898925
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,1,16,power_law_1.01,0.09488384246826172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,1,16,power_law_1.01,0.0913856029510498
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,1,16,power_law_1.01,0.04752511978149414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,1,16,power_law_1.01,1.7570343017578125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,1,16,power_law_1.01,0.09926655769348144
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,1,16,power_law_1.01,0.168591365814209
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,1,16,power_law_1.01,0.10315648078918457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,1,16,power_law_1.01,0.10070783615112304
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,1,16,power_law_1.01,0.09831808090209962
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,1,16,power_law_1.01,0.10005887985229492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,1,16,power_law_1.01,0.09792511940002442
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,1,16,power_law_1.01,0.10687487602233887
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,1,16,power_law_1.01,0.129486083984375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,1,16,power_law_1.01,0.1409331226348877
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,1,16,power_law_1.01,0.16528127670288087
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,1,16,power_law_1.01,2.3588748168945313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,1,16,power_law_1.01,0.26966783523559573
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,1,16,power_law_1.01,0.3268889617919922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,1,16,power_law_1.01,0.09389439582824707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,1,16,power_law_1.01,0.4475571060180664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,1,16,power_law_1.2,0.06994815826416015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,1,16,power_law_1.01,0.5590156936645508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,1,16,power_law_1.2,0.12181119918823242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,1,16,power_law_1.2,0.11331456184387208
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,1,16,power_law_1.2,0.12434559822082521
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,1,16,power_law_1.2,0.12192511558532715
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,1,16,power_law_1.01,0.9035711669921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,1,16,power_law_1.2,0.13936767578125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,1,16,power_law_1.2,0.1329766368865967
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,1,16,power_law_1.2,0.13940095901489258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,1,16,power_law_1.01,0.19957759857177734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,1,16,power_law_1.2,0.1572467231750488
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,1,16,power_law_1.2,0.16387840270996096
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,1,16,power_law_1.01,1.2016806030273437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,1,16,power_law_1.2,0.16640640258789063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,1,16,power_law_1.2,0.16792959213256836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,1,16,power_law_1.2,0.18519168853759765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,1,16,power_law_1.2,0.19252607345581055
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,1,16,power_law_1.2,0.22336767196655272
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,1,16,power_law_1.2,0.2787443161010742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,1,16,power_law_1.2,0.2952435111999512
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,1,16,power_law_1.2,0.11465855598449708
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,1,16,power_law_1.2,0.3533670425415039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,1,16,power_law_1.2,0.39958526611328127
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,1,16,power_law_1.2,0.5578841781616211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,1,16,power_law_1.2,0.7405683135986327
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,1,16,power_law_1.2,0.07818751811981202
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,1,16,power_law_1.2,1.1595059204101563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,1,16,power_law_1.2,0.07737472057342529
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,1,16,power_law_1.2,0.08046719551086426
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,1,16,power_law_1.2,1.5032051086425782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,1,16,power_law_1.2,0.0814361572265625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,1,16,power_law_1.2,0.0682099199295044
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,1,16,power_law_1.2,0.047528958320617674
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,1,16,power_law_1.2,0.08726655960083007
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,1,16,power_law_1.2,0.08849791526794434
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,1,16,power_law_1.2,0.09136639595031738
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,1,16,power_law_1.2,0.09967743873596191
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,1,16,power_law_1.2,0.09238016128540039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,1,16,power_law_1.2,2.2337983703613284
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,1,16,power_law_1.2,0.10213888168334961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,1,16,power_law_1.2,0.09859583854675294
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,1,16,power_law_1.2,0.09728896141052246
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,1,16,power_law_1.2,0.13418111801147461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,1,16,power_law_1.2,0.11024255752563476
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,1,16,power_law_1.2,0.10357119560241698
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,1,16,power_law_1.2,0.13288576126098633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,1,16,power_law_1.2,0.14337408065795898
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,1,16,power_law_1.2,0.1817740821838379
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,1,16,power_law_1.2,0.21129344940185546
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,1,16,power_law_1.2,0.2774886322021485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,1,16,power_law_1.2,0.3592140960693359
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,1,16,power_law_1.2,3.0538470458984373
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,2,8,balanced,0.022510080337524413
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,2,8,balanced,0.023509759902954102
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,2,8,balanced,0.03961215972900391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,2,8,balanced,0.04998271942138672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,2,8,balanced,0.07470848083496093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,1,16,power_law_1.2,0.684684829711914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,1,16,power_law_1.2,0.09997823715209961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,2,8,balanced,0.169749755859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,2,8,balanced,0.17303552627563476
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,2,8,balanced,0.17602943420410155
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,2,8,balanced,0.1818726348876953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,2,8,balanced,0.1732966423034668
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,1,16,power_law_1.2,1.0874495697021485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,2,8,balanced,0.1797439956665039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,2,8,balanced,0.1808076858520508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,2,8,balanced,0.18363391876220703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,2,8,balanced,0.19029375076293947
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,2,8,balanced,0.1950707244873047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,2,8,balanced,0.20225408554077146
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,2,8,balanced,0.22841472625732423
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,1,16,power_law_1.2,1.4022323608398437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,2,8,balanced,0.24724096298217774
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,2,8,balanced,0.28417152404785156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,2,8,balanced,0.3202483367919922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,2,8,balanced,0.3937420654296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,2,8,balanced,0.13018624305725096
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,2,8,balanced,0.4623104095458984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,2,8,balanced,0.024632320404052735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,2,8,balanced,0.025137920379638672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,2,8,balanced,0.02644223928451538
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,2,8,balanced,0.6503923034667969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,2,8,balanced,0.055024638175964355
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,2,8,balanced,0.8209420776367187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,2,8,balanced,0.08670080184936524
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,2,8,balanced,0.09910911560058594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,2,8,balanced,0.1019916820526123
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,2,8,balanced,0.10368767738342285
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,2,8,balanced,0.10546175956726074
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,2,8,balanced,0.0960588836669922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,1,16,power_law_1.2,0.5406220626831055
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,2,8,balanced,0.09706496238708497
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,2,8,balanced,0.09883392333984375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,2,8,balanced,0.10183168411254884
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,2,8,balanced,0.10751232147216797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,2,8,balanced,0.12066687583923338
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,2,8,balanced,0.13625599861145019
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,2,8,balanced,0.15033727645874023
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,2,8,balanced,1.5504614257812501
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,2,8,balanced,0.20410240173339841
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,2,8,balanced,0.03695231914520264
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,2,8,balanced,0.31808767318725584
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,2,8,balanced,0.4475545501708984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,2,8,balanced,0.5567808151245117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,2,8,balanced,1.1940006256103515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,2,8,power_law_1.01,0.048276481628417967
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,2,8,balanced,0.1112384033203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,2,8,power_law_1.01,0.07032703876495361
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,2,8,balanced,0.8126156616210938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,2,8,power_law_1.01,0.09581567764282226
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,2,8,balanced,0.17897983551025393
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,2,8,power_law_1.01,0.09722496032714843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,2,8,power_law_1.01,0.09531519889831544
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,2,8,power_law_1.01,0.14921343803405762
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,2,8,power_law_1.01,0.13272831916809083
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,2,8,balanced,1.0514240264892578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,2,8,power_law_1.01,0.1601523208618164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,2,8,power_law_1.01,0.14148608207702637
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,2,8,power_law_1.01,0.15101056098937987
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,2,8,power_law_1.01,0.1490163230895996
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,2,8,power_law_1.01,0.15989503860473633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,2,8,power_law_1.01,0.1545638370513916
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,2,8,power_law_1.01,0.16498559951782227
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,2,8,power_law_1.01,0.17269887924194335
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,2,8,power_law_1.01,0.19320575714111327
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,2,8,power_law_1.01,0.22380544662475588
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,2,8,power_law_1.01,0.2528179168701172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,2,8,power_law_1.01,0.08435456275939941
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,2,8,power_law_1.01,0.29443071365356444
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,2,8,power_law_1.01,0.3420108795166016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,2,8,balanced,0.26014720916748046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,2,8,power_law_1.01,0.4313049697875977
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,2,8,power_law_1.01,0.547462387084961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,2,8,power_law_1.01,0.0511359977722168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,2,8,power_law_1.01,0.7789043426513672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,2,8,power_law_1.01,0.058336000442504886
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,2,8,power_law_1.01,0.9645887756347656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,2,8,power_law_1.01,0.06425983905792236
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,2,8,power_law_1.01,0.0761958408355713
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,2,8,power_law_1.01,0.07221759796142578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,2,8,power_law_1.01,0.083056640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,2,8,power_law_1.01,0.09782655715942383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,2,8,power_law_1.01,1.4860365295410156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,2,8,power_law_1.01,0.09982208251953126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,2,8,power_law_1.01,0.10296192169189453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,2,8,power_law_1.01,0.03597824096679687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,2,8,power_law_1.01,0.10634112358093262
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,2,8,power_law_1.01,0.09626879692077636
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,2,8,power_law_1.01,0.09889535903930664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,2,8,power_law_1.01,0.08981375694274903
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,2,8,power_law_1.01,0.10269951820373535
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,2,8,power_law_1.01,1.9290573120117187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,2,8,power_law_1.01,0.10878335952758789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,2,8,power_law_1.01,0.12931327819824218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,2,8,power_law_1.01,0.13978879928588867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,2,8,power_law_1.01,0.16853759765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,2,8,power_law_1.01,0.2054604721069336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,2,8,power_law_1.01,0.24780031204223635
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,2,8,power_law_1.01,0.09630208015441895
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,2,8,power_law_1.2,0.048014078140258785
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,2,8,power_law_1.01,0.3197107124328613
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,2,8,power_law_1.01,0.4652096176147461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,2,8,power_law_1.2,0.06848896026611329
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,2,8,power_law_1.01,0.5582310485839843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,2,8,power_law_1.2,0.08781567573547364
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,2,8,power_law_1.2,0.09715968132019043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,2,8,power_law_1.2,0.10853504180908202
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,2,8,power_law_1.01,0.8289817810058594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,2,8,power_law_1.2,0.15274368286132814
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,2,8,power_law_1.2,0.1384012794494629
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,2,8,power_law_1.2,0.140380163192749
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,2,8,power_law_1.01,1.0106649780273438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,2,8,power_law_1.2,0.15169407844543456
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,2,8,power_law_1.2,0.16898303985595703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,2,8,power_law_1.2,0.15014271736145018
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,2,8,power_law_1.2,0.16719871520996094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,2,8,power_law_1.2,0.17733760833740236
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,2,8,power_law_1.2,0.185614070892334
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,2,8,power_law_1.2,0.20695936203002932
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,2,8,power_law_1.2,0.22685951232910156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,2,8,power_law_1.2,0.07002240180969238
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,2,8,power_law_1.2,0.27282304763793946
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,2,8,power_law_1.2,0.30240640640258787
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,2,8,power_law_1.2,0.3584832000732422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,2,8,power_law_1.2,0.4649740982055664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,2,8,power_law_1.2,0.14033920288085938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,2,8,power_law_1.2,0.03595776081085205
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,2,8,power_law_1.2,0.5899033737182617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,2,8,power_law_1.2,0.04966784000396728
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,2,8,power_law_1.2,0.05217919826507569
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,2,8,power_law_1.2,0.06389376163482666
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,2,8,power_law_1.2,0.07705088138580322
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,2,8,power_law_1.2,0.09332736015319824
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,2,8,power_law_1.2,1.0762214660644531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,2,8,power_law_1.2,0.09576319694519043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,2,8,power_law_1.2,0.09748096466064453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,2,8,power_law_1.2,0.09942272186279297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,2,8,power_law_1.2,0.09719039916992187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,2,8,power_law_1.2,0.09939711570739747
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,2,8,power_law_1.2,0.09663104057312012
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,2,8,power_law_1.2,0.09938559532165528
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,2,8,power_law_1.2,1.7016294860839842
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,2,8,power_law_1.2,0.10138112068176271
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,2,8,power_law_1.2,0.11348992347717286
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,2,8,power_law_1.2,0.12780927658081054
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,2,8,power_law_1.2,0.14604415893554687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,2,8,power_law_1.2,0.16978815078735351
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,2,8,power_law_1.2,0.0737395191192627
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,2,8,power_law_1.2,0.2111078453063965
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,2,8,power_law_1.2,2.3818981933593752
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,2,8,power_law_1.2,0.2775001525878906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,2,8,power_law_1.2,0.35099521636962894
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,4,4,balanced,0.020927999019622803
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,4,4,balanced,0.022437119483947755
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,2,8,power_law_1.2,0.4795673751831055
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,2,8,power_law_1.2,0.6166502380371094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,4,4,balanced,0.04895487785339355
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,2,8,power_law_1.2,0.8333414459228516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,4,4,balanced,0.07310976028442383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,4,4,balanced,0.12516863822937013
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,2,8,power_law_1.2,0.9171417236328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,4,4,balanced,0.1277888011932373
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,4,4,balanced,0.12857215881347656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,4,4,balanced,0.16162048339843751
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,4,4,balanced,0.16328960418701172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,4,4,balanced,0.16378623962402344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,4,4,balanced,0.1649087905883789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,2,8,power_law_1.2,1.2450931549072266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,2,8,power_law_1.2,0.10126208305358886
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,4,4,balanced,0.17112960815429687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,4,4,balanced,0.18083328247070313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,4,4,balanced,0.1848640060424805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,4,4,balanced,0.038585600852966306
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,4,4,balanced,0.2105356788635254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,4,4,balanced,0.22474239349365233
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,4,4,balanced,0.2702412796020508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,4,4,balanced,0.3037721633911133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,4,4,balanced,0.37893760681152344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,4,4,balanced,0.44573696136474605
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,4,4,balanced,0.6245542526245117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,4,4,balanced,0.16692224502563477
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,4,4,balanced,0.023486719131469727
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,4,4,balanced,0.790909423828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,4,4,balanced,0.03947263956069946
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,4,4,balanced,0.19078784942626953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,4,4,balanced,0.05378943920135498
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,4,4,balanced,0.08911231994628907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,4,4,balanced,0.08840448379516601
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,4,4,balanced,0.08993791580200196
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,4,4,balanced,1.1517721557617189
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,4,4,balanced,0.10269696235656738
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,4,4,balanced,0.10306431770324706
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,4,4,balanced,0.10490367889404298
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,4,4,balanced,0.10645631790161132
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,4,4,balanced,0.1067840003967285
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,4,4,balanced,0.11239808082580567
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,4,4,balanced,1.4892057800292968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,4,4,balanced,0.11587967872619628
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,4,4,balanced,0.1232038402557373
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,4,4,balanced,0.1389414405822754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,4,4,balanced,0.1481689643859863
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,4,4,balanced,0.1753113555908203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,4,4,balanced,0.2479449653625488
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,4,4,balanced,0.29817728042602537
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,4,4,balanced,0.026849279403686522
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,4,4,balanced,0.4315929412841797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,4,4,power_law_1.01,0.037620480060577395
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,4,4,balanced,0.5293939208984375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,4,4,power_law_1.01,0.047864317893981934
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,4,4,power_law_1.01,0.05609216213226318
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,4,4,balanced,0.7716761779785156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,4,4,power_law_1.01,0.06820223808288575
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,4,4,power_law_1.01,0.08157567977905274
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,4,4,power_law_1.01,0.0794547176361084
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,4,4,balanced,0.2012339210510254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,4,4,balanced,0.9996377563476562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,4,4,power_law_1.01,0.09937151908874511
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,4,4,power_law_1.01,0.11236096382141114
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,4,4,power_law_1.01,0.12240384101867677
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,4,4,power_law_1.01,0.12495231628417969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,4,4,power_law_1.01,0.14417407989501954
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,4,4,power_law_1.01,0.13755904197692873
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,4,4,power_law_1.01,0.14899328231811523
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,4,4,power_law_1.01,0.1519808006286621
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,4,4,power_law_1.01,0.18210559844970703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,4,4,power_law_1.01,0.20394367218017578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,4,4,balanced,0.10978176116943358
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,4,4,power_law_1.01,0.22773887634277346
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,4,4,balanced,0.023060479164123536
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,4,4,power_law_1.01,0.2870425605773926
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,4,4,power_law_1.01,0.3265139389038086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,4,4,power_law_1.01,0.3994764709472656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,4,4,power_law_1.01,0.5146252822875976
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,4,4,power_law_1.01,0.024728319644927978
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,4,4,power_law_1.01,0.1467852783203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,4,4,power_law_1.01,0.69404541015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,4,4,power_law_1.01,0.03769727945327759
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,4,4,power_law_1.01,0.04247424125671387
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,4,4,power_law_1.01,0.050682878494262694
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,4,4,power_law_1.01,0.9248000335693358
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,4,4,power_law_1.01,0.059552001953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,4,4,power_law_1.01,0.06023808002471924
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,4,4,power_law_1.01,0.08030847549438477
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,4,4,power_law_1.01,0.07936511993408203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,4,4,power_law_1.01,0.08687487602233887
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,4,4,power_law_1.01,0.09697279930114747
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,4,4,power_law_1.01,0.09925120353698731
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,4,4,power_law_1.01,1.2971661376953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,4,4,power_law_1.01,0.10140031814575194
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,4,4,power_law_1.01,0.10265472412109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,4,4,power_law_1.01,0.10759936332702638
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,4,4,power_law_1.01,0.10951935768127441
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,4,4,power_law_1.01,0.11673088073730468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,4,4,power_law_1.01,0.1414796829223633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,4,4,power_law_1.01,0.13413503646850586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,4,4,power_law_1.01,0.14701184272766113
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,4,4,power_law_1.01,1.7357478332519531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,4,4,power_law_1.01,0.2061260795593262
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,4,4,power_law_1.01,0.2706291198730469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,4,4,power_law_1.01,0.3282726287841797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,4,4,power_law_1.01,0.45432575225830074
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,4,4,power_law_1.2,0.03737087965011597
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,4,4,power_law_1.01,0.5924838256835938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,4,4,power_law_1.01,0.10548992156982422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,4,4,power_law_1.2,0.047504639625549315
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,4,4,power_law_1.2,0.05330175876617431
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,4,4,power_law_1.01,0.8390553283691407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,4,4,power_law_1.2,0.06669439792633056
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,4,4,power_law_1.2,0.07338111877441407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,4,4,power_law_1.2,0.11613439559936523
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,4,4,power_law_1.2,0.10961536407470703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,4,4,power_law_1.2,0.1212172794342041
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,4,4,power_law_1.01,1.0923788452148437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,4,4,power_law_1.2,0.12502400398254393
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,4,4,power_law_1.2,0.13055232048034668
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,4,4,power_law_1.2,0.1354956817626953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,4,4,power_law_1.2,0.13852160453796386
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,4,4,power_law_1.2,0.15314687728881834
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,4,4,power_law_1.2,0.1622271919250488
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,4,4,power_law_1.2,0.1908940887451172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,4,4,power_law_1.2,0.21856639862060545
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,4,4,power_law_1.2,0.25775615692138676
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,4,4,power_law_1.2,0.2980300712585449
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,4,4,power_law_1.01,0.17271039962768556
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,4,4,power_law_1.2,0.3386304092407227
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,4,4,power_law_1.2,0.4369023895263672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,4,4,power_law_1.2,0.542426872253418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,4,4,power_law_1.2,0.024555521011352537
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,4,4,power_law_1.2,0.14965120315551758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,4,4,power_law_1.2,0.7407769775390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,4,4,power_law_1.2,0.036852478981018066
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,4,4,power_law_1.2,0.038972160816192626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,4,4,power_law_1.2,0.04770431995391845
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,4,4,power_law_1.2,1.0075353240966796
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,4,4,power_law_1.2,0.07968128204345704
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,4,4,power_law_1.2,0.07739391803741455
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,4,4,power_law_1.2,0.08629759788513183
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,4,4,power_law_1.2,0.09446528434753418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,4,4,power_law_1.2,0.09939840316772461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,4,4,power_law_1.2,0.10122112274169921
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,4,4,power_law_1.2,0.10215807914733888
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,4,4,power_law_1.2,0.10542207717895509
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,4,4,power_law_1.2,0.10948096275329591
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,4,4,power_law_1.2,1.8177561950683594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,4,4,power_law_1.2,0.11048064231872559
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,4,4,power_law_1.2,0.11993984222412109
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,4,4,power_law_1.2,0.13783295631408693
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,4,4,power_law_1.2,0.056526079177856445
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,4,4,power_law_1.2,0.06470272064208984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,4,4,power_law_1.2,0.15135616302490235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,4,4,power_law_1.2,0.1827084732055664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,4,4,power_law_1.2,0.22030208587646483
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,4,4,power_law_1.2,1.3735769653320313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,4,4,power_law_1.2,0.3419852828979492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,8,2,balanced,0.020944640636444092
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,4,4,power_law_1.2,0.4828620910644531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,8,2,balanced,0.032668159008026124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,4,4,power_law_1.2,0.6185766220092773
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,8,2,balanced,0.04727935791015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,8,2,balanced,0.07244160175323486
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,8,2,balanced,0.12655743598937988
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,8,2,balanced,0.13282048225402832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,4,4,power_law_1.2,0.8790451049804687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,8,2,balanced,0.12481663703918458
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,4,4,power_law_1.2,0.08239104270935058
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,8,2,balanced,0.12680831909179688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,8,2,balanced,0.1312012767791748
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,8,2,balanced,0.13636863708496094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,8,2,balanced,0.13897727966308593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,8,2,balanced,0.13946751594543455
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,4,4,power_law_1.2,1.126872329711914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,8,2,balanced,0.1455884838104248
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,8,2,balanced,0.14937600135803222
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,8,2,balanced,0.15653504371643065
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,8,2,balanced,0.16784255981445312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,8,2,balanced,0.029542400836944583
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,8,2,balanced,0.23256448745727537
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,8,2,balanced,0.20956159591674806
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,8,2,balanced,0.2993395233154297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,8,2,balanced,0.34718334197998046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,8,2,balanced,0.5022681427001954
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,8,2,balanced,0.020624639987945555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,8,2,balanced,0.6024486541748046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,8,2,balanced,0.021984000205993653
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,8,2,balanced,0.14402303695678711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,4,4,power_law_1.2,0.2855641555786133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,8,2,balanced,0.036514561176300045
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,8,2,balanced,0.8837875366210938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,8,2,balanced,0.08489727973937988
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,8,2,balanced,0.08586624145507812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,8,2,balanced,0.08673279762268067
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,8,2,balanced,1.1165990447998047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,8,2,balanced,0.08816512107849121
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,8,2,balanced,0.09053695678710938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,8,2,balanced,0.09238016128540039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,8,2,balanced,0.09838463783264159
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,8,2,balanced,0.1033625602722168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,8,2,balanced,0.10723199844360351
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,8,2,balanced,0.1125107192993164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,8,2,balanced,0.12515968322753906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,8,2,balanced,0.13491711616516114
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,8,2,balanced,0.22928768157958984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,8,2,balanced,0.19571968078613283
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,8,2,balanced,0.25946495056152347
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,8,2,balanced,0.29310848236083986
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,8,2,balanced,0.4343692779541016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,8,2,balanced,0.08344191551208496
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,8,2,balanced,0.5248972702026367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,8,2,balanced,0.0867955207824707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,8,2,power_law_1.01,0.02192255973815918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,8,2,power_law_1.01,0.03865983963012695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,8,2,balanced,0.7737522888183593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,8,2,balanced,1.0040191650390624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,8,2,power_law_1.01,0.07799551963806153
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,8,2,power_law_1.01,0.10923392295837402
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,8,2,power_law_1.01,0.10038016319274903
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,8,2,power_law_1.01,0.1007372760772705
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,8,2,power_law_1.01,0.10266112327575683
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,8,2,balanced,0.025173759460449217
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,8,2,power_law_1.01,0.11883008003234863
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,8,2,balanced,0.054402561187744135
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,8,2,power_law_1.01,0.1256499195098877
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,8,2,power_law_1.01,0.13455615997314455
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,8,2,power_law_1.01,0.15036928176879882
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,8,2,power_law_1.01,0.14420991897583008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,8,2,power_law_1.01,0.17853952407836915
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,8,2,power_law_1.01,0.2037606430053711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,8,2,power_law_1.01,0.2665535926818848
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,8,2,power_law_1.01,0.275283203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,8,2,power_law_1.01,0.1261516761779785
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,8,2,power_law_1.01,0.34596607208251956
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,8,2,power_law_1.01,0.05173632144927979
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,8,2,power_law_1.01,0.5470745468139648
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,8,2,balanced,0.18531967163085938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,8,2,power_law_1.01,0.6925068664550781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,8,2,power_law_1.01,0.03447168111801148
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,8,2,power_law_1.01,0.022664959430694583
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,8,2,power_law_1.01,0.032309761047363286
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,8,2,power_law_1.01,0.957770233154297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,8,2,power_law_1.01,0.06558720111846925
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,8,2,power_law_1.01,0.049625601768493656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,8,2,power_law_1.01,0.05609471797943115
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,8,2,power_law_1.01,1.222899169921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,8,2,power_law_1.01,0.07071231842041016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,8,2,power_law_1.01,0.07274752140045167
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,8,2,power_law_1.01,0.07754752159118652
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,8,2,power_law_1.01,0.07150335788726807
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,8,2,power_law_1.01,0.08094592094421386
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,8,2,power_law_1.01,0.08391807556152345
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,8,2,power_law_1.01,0.09313535690307617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,8,2,power_law_1.01,0.041164798736572264
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,8,2,power_law_1.01,0.09823616027832031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,8,2,power_law_1.01,0.10890624046325684
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,8,2,power_law_1.01,0.11476479530334474
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,8,2,power_law_1.01,0.10584063529968261
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,8,2,power_law_1.01,0.1306828784942627
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,8,2,power_law_1.01,0.1440998363494873
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,8,2,power_law_1.01,0.18905216217041015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,8,2,power_law_1.01,0.3400409698486328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,8,2,power_law_1.01,0.4667379379272461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,8,2,power_law_1.01,0.5874995040893555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,8,2,power_law_1.2,0.02186624050140381
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,8,2,power_law_1.01,0.07052671909332275
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,8,2,power_law_1.2,0.03324543952941895
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,8,2,power_law_1.2,0.03615744113922119
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,8,2,power_law_1.01,1.087426528930664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,8,2,power_law_1.2,0.06098688125610352
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,8,2,power_law_1.2,0.07423744201660157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,8,2,power_law_1.2,0.1118899154663086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,8,2,power_law_1.2,0.09994367599487305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,8,2,power_law_1.2,0.09616640090942383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,8,2,power_law_1.2,0.10301568031311034
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,8,2,power_law_1.2,0.11580287933349609
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,8,2,power_law_1.01,0.41040126800537113
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,8,2,power_law_1.2,0.11974399566650391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,8,2,power_law_1.01,0.02165760040283203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,8,2,power_law_1.2,0.1403660774230957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,8,2,power_law_1.2,0.15470463752746583
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,8,2,power_law_1.2,0.1829939270019531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,8,2,power_law_1.2,0.2327564811706543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,8,2,power_law_1.2,0.12978048324584962
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,8,2,power_law_1.2,0.26850175857543945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,8,2,power_law_1.2,0.2844710350036621
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,8,2,power_law_1.2,0.3567027282714844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,8,2,power_law_1.2,0.41796863555908204
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,8,2,power_law_1.01,0.2890291213989258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,8,2,power_law_1.01,0.21518976211547852
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,8,2,power_law_1.2,0.5747238540649414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,8,2,power_law_1.2,0.7082444763183593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,8,2,power_law_1.2,0.021299200057983397
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,8,2,power_law_1.2,0.049492478370666504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,8,2,power_law_1.2,0.02371455907821655
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,8,2,power_law_1.2,0.031354880332946776
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,8,2,power_law_1.2,1.0004045104980468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,8,2,power_law_1.2,0.03914495944976807
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,8,2,power_law_1.2,0.04882815837860107
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,8,2,power_law_1.2,1.2908248901367188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,8,2,power_law_1.2,0.06991104125976563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,8,2,power_law_1.2,0.07258624076843262
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,8,2,power_law_1.2,0.068472318649292
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,8,2,power_law_1.01,0.8377394866943358
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,8,2,power_law_1.2,0.07818111896514893
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,8,2,power_law_1.2,0.08204799652099609
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,8,2,power_law_1.2,0.08502143859863281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,8,2,power_law_1.2,0.126179838180542
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,8,2,power_law_1.2,0.09518976211547851
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,8,2,power_law_1.2,0.0999846363067627
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,8,2,power_law_1.2,0.05256959915161132
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,8,2,power_law_1.2,0.13559679985046386
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,8,2,power_law_1.2,0.14911231994628907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,8,2,power_law_1.2,0.2184908866882324
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,8,2,power_law_1.2,0.07338367938995362
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,8,2,power_law_1.2,0.3498534393310547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,8,2,power_law_1.2,0.4774220657348633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,8,2,power_law_1.2,0.161658878326416
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,8,2,power_law_1.2,0.5993510437011719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,16,1,balanced,0.029013760089874267
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,16,1,balanced,0.03970880031585693
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,8,2,power_law_1.2,0.8567155456542969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,16,1,balanced,0.04972544193267822
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,16,1,balanced,0.06417471885681152
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,16,1,balanced,0.0928166389465332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,16,1,balanced,0.16332288742065432
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,16,1,balanced,0.1626361656188965
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,16,1,balanced,0.16350400924682618
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,16,1,balanced,0.1642790412902832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,16,1,balanced,0.16589183807373048
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,8,2,power_law_1.2,1.0918681335449218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,16,1,balanced,0.1685856056213379
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,16,1,balanced,0.17156543731689453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,16,1,balanced,0.1742585563659668
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,16,1,balanced,0.1791276741027832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,16,1,balanced,0.19049280166625976
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,16,1,balanced,0.19153343200683595
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,16,1,balanced,0.19634880065917967
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,16,1,balanced,0.20973567962646483
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,16,1,balanced,0.2334432029724121
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,16,1,balanced,0.2668441581726074
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,16,1,balanced,0.3010323143005371
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,16,1,balanced,0.3978752136230469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,16,1,balanced,0.4401926422119141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,16,1,balanced,0.6249836730957031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,8,2,power_law_1.2,0.1119961643218994
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,8,2,power_law_1.2,0.1132748794555664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,16,1,balanced,0.030864639282226564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,16,1,balanced,0.8192889404296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,16,1,balanced,0.03412607908248901
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,16,1,balanced,0.040202879905700685
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,16,1,balanced,0.05857600212097168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,16,1,balanced,0.0926790428161621
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,16,1,balanced,0.14888192176818849
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,16,1,balanced,0.15091263771057128
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,16,1,balanced,0.15157567977905273
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,16,1,balanced,0.15872896194458008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,16,1,balanced,1.1860511779785157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,16,1,balanced,0.15295552253723144
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,8,2,power_law_1.2,0.2936742401123047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,16,1,balanced,0.15519424438476564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,16,1,balanced,0.1566528034210205
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,16,1,balanced,0.15897855758666993
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,16,1,balanced,0.16062335968017577
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,16,1,balanced,0.19007488250732424
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,16,1,balanced,0.19238847732543946
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,16,1,balanced,0.19637632369995117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,16,1,balanced,0.20471359252929688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,16,1,balanced,1.5717837524414062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,16,1,balanced,0.21212095260620117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,16,1,balanced,0.23201215744018552
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,16,1,balanced,0.25023231506347654
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,16,1,balanced,0.2912563133239746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,8,2,power_law_1.2,0.19573120117187498
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,16,1,balanced,0.3307372665405274
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,16,1,power_law_1.01,0.028592000007629393
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,16,1,balanced,0.521075210571289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,16,1,power_law_1.01,0.03586944103240967
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,16,1,power_law_1.01,0.06380544185638429
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,16,1,balanced,0.6047993469238281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,16,1,power_law_1.01,0.07988224029541016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,16,1,power_law_1.01,0.1302291202545166
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,16,1,power_law_1.01,0.13830719947814943
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,16,1,balanced,0.8876038360595704
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,16,1,power_law_1.01,0.13982784271240234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,16,1,power_law_1.01,0.14230527877807617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,16,1,power_law_1.01,0.1505190372467041
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,16,1,balanced,1.1708940887451171
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,16,1,power_law_1.01,0.15785792350769043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,16,1,power_law_1.01,0.18162239074707032
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,16,1,power_law_1.01,0.16278783798217772
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,16,1,power_law_1.01,0.1650214385986328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,16,1,power_law_1.01,0.04948544025421143
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,16,1,power_law_1.01,0.16823999404907225
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,16,1,power_law_1.01,0.1819264030456543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,16,1,power_law_1.01,0.23451135635375975
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,16,1,power_law_1.01,0.10474944114685059
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,16,1,power_law_1.01,0.33093120574951174
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,16,1,power_law_1.01,0.3907487869262695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,16,1,power_law_1.01,0.47805057525634764
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,16,1,power_law_1.01,0.6567565155029297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,16,1,power_law_1.01,0.031108479499816894
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,16,1,power_law_1.01,0.20473407745361327
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,16,1,power_law_1.01,0.841244125366211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,16,1,power_law_1.01,0.30488960266113285
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,16,1,power_law_1.01,0.03439232110977173
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,16,1,power_law_1.01,0.04280576229095459
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,16,1,power_law_1.01,0.05791615962982177
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,16,1,power_law_1.01,0.07622144222259522
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,16,1,power_law_1.01,0.12193856239318848
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,16,1,power_law_1.01,0.09192831993103028
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,16,1,power_law_1.01,0.1263596820831299
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,16,1,power_law_1.01,1.2089894104003907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,16,1,power_law_1.01,0.12828800201416016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,16,1,power_law_1.01,0.13076224327087402
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,16,1,power_law_1.01,0.13540032386779785
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,16,1,power_law_1.01,0.1385587215423584
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,16,1,power_law_1.01,0.1441759967803955
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,16,1,power_law_1.01,0.17550399780273437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,16,1,power_law_1.01,1.5833952331542969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,16,1,power_law_1.01,0.1768671989440918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,16,1,power_law_1.01,0.18593151092529298
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,16,1,power_law_1.01,0.19814720153808593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,16,1,power_law_1.01,0.21678848266601564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,16,1,power_law_1.01,0.25813568115234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,16,1,power_law_1.01,0.2853055953979492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,16,1,power_law_1.01,0.35076160430908204
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,16,1,power_law_1.01,0.42671039581298825
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,16,1,power_law_1.01,0.14075648307800293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,16,1,power_law_1.01,0.5593535995483399
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,16,1,power_law_1.2,0.03828927993774414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,16,1,power_law_1.01,0.7106201934814453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,16,1,power_law_1.2,0.0763481616973877
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,16,1,power_law_1.2,0.1018956756591797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,16,1,power_law_1.01,0.9941158294677734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,16,1,power_law_1.2,0.12763327598571778
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,16,1,power_law_1.2,0.13521280288696289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,16,1,power_law_1.2,0.13922112464904784
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,16,1,power_law_1.01,1.2661875152587891
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,16,1,power_law_1.2,0.1502988815307617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,16,1,power_law_1.2,0.061272959709167484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,16,1,power_law_1.2,0.15742143630981445
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,16,1,power_law_1.2,0.1687513542175293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,16,1,power_law_1.2,0.02783679962158203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,16,1,power_law_1.2,0.17555391311645507
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,16,1,power_law_1.2,0.04988351821899414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,16,1,power_law_1.2,0.18701824188232422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,16,1,power_law_1.2,0.21677120208740233
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,16,1,power_law_1.2,0.2434976005554199
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,16,1,power_law_1.2,0.14308671951293944
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,16,1,power_law_1.2,0.3176019287109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,16,1,power_law_1.2,0.3387609481811523
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,16,1,power_law_1.2,0.4023404693603515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,16,1,power_law_1.2,0.16399040222167968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,16,1,power_law_1.2,0.18739328384399415
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,16,1,power_law_1.2,0.4919340896606445
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,16,1,power_law_1.2,0.030956799983978274
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,16,1,power_law_1.2,0.03451200008392334
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,16,1,power_law_1.2,0.04293312072753906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,16,1,power_law_1.2,0.05646527767181396
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,16,1,power_law_1.2,0.8571302032470702
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,16,1,power_law_1.2,0.0912723159790039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,16,1,power_law_1.2,0.11983424186706544
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,16,1,power_law_1.2,0.12474495887756347
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,16,1,power_law_1.2,0.12712127685546876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,16,1,power_law_1.2,0.1303647994995117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,16,1,power_law_1.2,0.13328831672668456
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,16,1,power_law_1.2,1.2297535705566407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,16,1,power_law_1.2,0.13762816429138183
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,16,1,power_law_1.2,0.6677350616455078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,16,1,power_law_1.2,0.1409708786010742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,16,1,power_law_1.2,0.1457913589477539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,16,1,power_law_1.2,0.17714048385620118
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,16,1,power_law_1.2,0.1826572799682617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,16,1,power_law_1.2,1.6185093688964844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,16,1,power_law_1.2,0.188986873626709
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,16,1,power_law_1.2,0.07154560089111328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,16,1,power_law_1.2,0.20960704803466798
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,16,1,power_law_1.2,0.23475072860717772
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,16,1,power_law_1.2,0.2670566368103027
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,16,1,power_law_1.2,0.29413312911987305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,1,16,balanced,0.023646719455718994
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,16,1,power_law_1.2,0.36323326110839843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,1,16,balanced,0.0235532808303833
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,16,1,power_law_1.2,0.445016975402832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,1,16,balanced,0.042707200050354006
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,1,16,balanced,0.052255997657775874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,1,16,balanced,0.06200448036193847
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,16,1,power_law_1.2,0.5683283233642579
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,1,16,balanced,0.10385279655456543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,1,16,balanced,0.1016268825531006
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,1,16,balanced,0.10243328094482422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,1,16,balanced,0.10692480087280273
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,1,16,balanced,0.10718591690063477
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,1,16,balanced,0.11057791709899903
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,1,16,balanced,0.11029248237609864
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,1,16,balanced,0.11262207984924318
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,1,16,balanced,0.1221132755279541
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,1,16,balanced,0.1251968002319336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,1,16,balanced,0.11672703742980957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,1,16,balanced,0.11938303947448729
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,16,1,power_law_1.2,1.0048191833496094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,1,16,balanced,0.13488127708435058
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,1,16,balanced,0.16515712738037108
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,1,16,balanced,0.1781337547302246
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,1,16,balanced,0.2505843162536621
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,1,16,balanced,0.2709811210632324
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,16,1,power_law_1.2,1.2914009094238281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,1,16,balanced,0.3877452850341797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,1,16,balanced,0.10231040000915528
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,1,16,balanced,0.47633022308349615
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,16,1,power_law_1.2,0.7208934020996094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,1,16,power_law_1.01,0.07268735885620117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,1,16,power_law_1.01,0.06648831844329835
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,1,16,balanced,0.6904013061523437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,1,16,power_law_1.01,0.0953996753692627
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,1,16,power_law_1.01,0.09412351608276368
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,1,16,balanced,0.8950323486328126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,1,16,power_law_1.01,0.10144895553588866
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,1,16,power_law_1.01,0.10477952003479003
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,1,16,power_law_1.01,0.1074240016937256
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,1,16,power_law_1.01,0.10990847587585448
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,1,16,power_law_1.01,0.11124223709106444
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,1,16,power_law_1.01,0.11398271560668945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,1,16,power_law_1.01,0.12229375839233399
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,1,16,power_law_1.01,0.12516736030578612
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,1,16,power_law_1.01,0.1288486385345459
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,1,16,power_law_1.01,0.13422847747802735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,1,16,power_law_1.01,0.061479678153991704
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,1,16,power_law_1.01,0.16660608291625975
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,1,16,power_law_1.01,0.07203072071075439
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,1,16,power_law_1.01,0.20185855865478514
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,1,16,power_law_1.01,0.24445568084716798
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,1,16,power_law_1.01,0.30018432617187496
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,1,16,power_law_1.01,0.42191486358642577
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,1,16,power_law_1.01,0.10326144218444824
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,1,16,power_law_1.01,0.4869529724121094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,1,16,power_law_1.2,0.06138879776000976
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,1,16,power_law_1.01,0.7077977752685547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,1,16,power_law_1.2,0.07007743835449218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,1,16,power_law_1.2,0.08301312446594238
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,1,16,power_law_1.2,0.09671680450439453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,1,16,power_law_1.01,0.9530828857421876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,1,16,power_law_1.2,0.09778559684753418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,1,16,power_law_1.2,0.10345984458923341
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,1,16,power_law_1.2,0.10626815795898437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,1,16,power_law_1.2,0.10780544281005859
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,1,16,power_law_1.2,0.10950143814086914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,1,16,power_law_1.2,0.11479295730590819
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,1,16,power_law_1.2,0.12155391693115233
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,1,16,power_law_1.2,0.12318464279174805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,1,16,power_law_1.2,0.13011199951171876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,1,16,power_law_1.2,0.13106304168701172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,1,16,power_law_1.2,0.1479206371307373
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,1,16,power_law_1.2,0.07252736091613769
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,1,16,power_law_1.2,0.18940671920776367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,1,16,power_law_1.2,0.20792831420898436
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,1,16,power_law_1.2,0.2732774353027344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,1,16,power_law_1.2,0.351091194152832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,1,16,power_law_1.2,0.4487897491455078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,1,16,power_law_1.2,0.6629503631591798
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,1,16,power_law_1.01,1.4487680053710936
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,2,8,balanced,0.02101759910583496
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,2,8,balanced,0.02212224006652832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,2,8,balanced,0.03431040048599243
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,1,16,power_law_1.2,0.9015026855468751
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,2,8,balanced,0.047457280158996584
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,2,8,balanced,0.06414207935333252
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,1,16,power_law_1.01,1.9309849548339844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,1,16,power_law_1.2,1.2117491149902344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,2,8,balanced,0.10078975677490234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,2,8,balanced,0.09655679702758789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,2,8,balanced,0.10173312187194825
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,2,8,balanced,0.10403072357177734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,2,8,balanced,0.1035264015197754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,2,8,balanced,0.10698880195617674
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,2,8,balanced,0.11406335830688477
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,2,8,balanced,0.11668607711791992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,2,8,balanced,0.10732288360595703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,2,8,balanced,0.11133312225341796
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,2,8,balanced,0.10889087677001955
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,2,8,balanced,0.1125312042236328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,2,8,balanced,0.13025919914245604
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,2,8,balanced,0.13924480438232423
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,2,8,balanced,0.1849228858947754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,2,8,balanced,0.20451711654663085
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,1,16,power_law_1.2,0.0717580795288086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,2,8,balanced,0.2768614387512207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,2,8,balanced,0.31161088943481446
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,1,16,power_law_1.2,2.0685093688964846
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,2,8,power_law_1.01,0.04264832019805908
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,2,8,balanced,0.44446720123291017
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,2,8,power_law_1.01,0.059677438735961916
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,2,8,power_law_1.01,0.0530291223526001
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,2,8,balanced,0.5438489532470703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,2,8,power_law_1.01,0.0570636796951294
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,2,8,power_law_1.01,0.08738559722900391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,2,8,power_law_1.01,0.09512319564819335
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,2,8,balanced,0.7956082916259766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,2,8,power_law_1.01,0.09730815887451172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,1,16,power_law_1.2,2.467498168945313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,2,8,power_law_1.01,0.0984985637664795
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,2,8,power_law_1.01,0.10134655952453613
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,2,8,balanced,1.0378585815429688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,2,8,power_law_1.01,0.10538496017456053
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,2,8,power_law_1.01,0.11096192359924317
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,2,8,power_law_1.01,0.10758399963378906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,2,8,power_law_1.01,0.10861439704895019
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,2,8,power_law_1.01,0.11661439895629884
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,2,8,power_law_1.01,0.14566911697387694
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,2,8,power_law_1.01,0.17442047119140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,2,8,power_law_1.01,0.2162688064575195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,2,8,power_law_1.01,0.26073856353759767
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,2,8,power_law_1.01,0.3405260848999023
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,2,8,power_law_1.01,0.09552639961242676
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,2,8,power_law_1.01,0.42485504150390624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,2,8,power_law_1.01,0.6246271896362304
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,2,8,power_law_1.01,0.06591872215270997
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,2,8,power_law_1.01,0.1102412796020508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,2,8,power_law_1.01,0.7629977416992187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,2,8,power_law_1.2,0.05258368015289307
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,2,8,power_law_1.2,0.053052158355712885
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,2,8,power_law_1.2,0.052183041572570796
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,2,8,power_law_1.2,0.08912384033203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,2,8,power_law_1.2,0.09106047630310059
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,2,8,power_law_1.01,1.1112550354003905
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,2,8,power_law_1.2,0.09475199699401855
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,2,8,power_law_1.2,0.0965452766418457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,2,8,power_law_1.2,0.0982374382019043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,2,8,power_law_1.2,0.10314880371093751
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,2,8,power_law_1.2,0.10962944030761719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,2,8,power_law_1.2,0.10801024436950683
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,2,8,power_law_1.2,0.10684800148010254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,2,8,power_law_1.2,0.042639360427856446
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,2,8,power_law_1.2,0.10904064178466796
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,2,8,power_law_1.2,0.11688960075378418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,2,8,power_law_1.2,0.1317593574523926
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,2,8,power_law_1.2,0.1638515281677246
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,2,8,power_law_1.2,0.1753971290588379
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,2,8,power_law_1.2,0.23250303268432618
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,2,8,power_law_1.2,0.2817612838745117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,2,8,power_law_1.2,0.418073616027832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,2,8,power_law_1.2,0.47655166625976564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,4,4,balanced,0.020710399150848387
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,4,4,balanced,0.02192512035369873
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,4,4,balanced,0.033786880970001223
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,2,8,power_law_1.2,0.058073601722717284
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,4,4,balanced,0.0467955207824707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,2,8,power_law_1.01,1.5597325134277344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,4,4,balanced,0.06513535976409912
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,4,4,balanced,0.0791756820678711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,2,8,power_law_1.2,0.897991714477539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,4,4,balanced,0.10750080108642579
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,4,4,balanced,0.11066495895385742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,4,4,balanced,0.10932864189147948
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,4,4,balanced,0.11264127731323241
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,4,4,balanced,0.11865599632263184
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,4,4,balanced,0.1220032024383545
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,4,4,balanced,0.11028479576110839
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,4,4,balanced,0.11454079627990724
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,4,4,balanced,0.11169407844543458
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,4,4,balanced,0.11439359664916993
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,4,4,balanced,0.11799296379089355
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,4,4,balanced,0.12930303573608398
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,4,4,balanced,0.13731200218200684
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,4,4,balanced,0.18870655059814453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,4,4,balanced,0.2103232002258301
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,4,4,balanced,0.29984384536743164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,2,8,power_law_1.2,0.7445964813232422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,4,4,balanced,0.34254081726074215
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,2,8,power_law_1.2,1.83849853515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,4,4,balanced,0.49820289611816404
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,4,4,power_law_1.01,0.029164800643920897
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,4,4,power_law_1.01,0.042938880920410157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,4,4,balanced,0.6236774444580078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,4,4,power_law_1.01,0.05165567874908448
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,4,4,power_law_1.01,0.05966720104217529
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,4,4,power_law_1.01,0.07029376029968262
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,4,4,power_law_1.01,0.09203455924987794
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,4,4,balanced,0.911746597290039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,4,4,power_law_1.01,0.10013055801391602
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,4,4,power_law_1.01,0.10214655876159669
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,2,8,power_law_1.2,1.4620069885253906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,4,4,power_law_1.01,0.10528512001037597
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,4,4,power_law_1.01,0.11119744300842285
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,4,4,power_law_1.01,0.1135923194885254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,4,4,power_law_1.01,0.10767359733581543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,4,4,balanced,1.1883827209472657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,4,4,power_law_1.01,0.1106879997253418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,4,4,power_law_1.01,0.11433600425720214
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,4,4,power_law_1.01,0.13635456085205078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,4,4,power_law_1.01,0.15914879798889162
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,4,4,power_law_1.01,0.19426048278808594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,4,4,power_law_1.01,0.24469888687133787
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,4,4,power_law_1.01,0.043589119911193845
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,4,4,power_law_1.01,0.3079948806762695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,4,4,power_law_1.01,0.4080460739135742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,4,4,power_law_1.01,0.5596659088134766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,4,4,power_law_1.01,0.1067136001586914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,4,4,power_law_1.01,0.7045836639404297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,4,4,power_law_1.2,0.04068352222442627
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,4,4,power_law_1.01,0.10937343597412111
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,4,4,power_law_1.2,0.043828477859497075
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,4,4,power_law_1.2,0.051687679290771484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,4,4,power_law_1.2,0.05191296100616455
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,4,4,power_law_1.01,1.0068109130859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,4,4,power_law_1.2,0.0911411190032959
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,4,4,power_law_1.2,0.09824895858764648
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,4,4,power_law_1.2,0.10216064453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,4,4,power_law_1.2,0.10494976043701172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,4,4,power_law_1.2,0.11223936080932617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,4,4,power_law_1.2,0.11542655944824218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,4,4,power_law_1.2,0.10666751861572266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,4,4,power_law_1.01,1.5196287536621094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,4,4,power_law_1.2,0.11131775856018067
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,4,4,power_law_1.2,0.1096448040008545
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,4,4,power_law_1.2,0.11325695991516113
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,4,4,power_law_1.2,0.12071295738220215
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,4,4,power_law_1.2,0.16316543579101564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,4,4,power_law_1.2,0.145863676071167
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,4,4,power_law_1.2,0.20794752120971677
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,4,4,power_law_1.2,0.25285503387451175
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,4,4,power_law_1.2,0.03139967918395996
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,4,4,power_law_1.2,0.3518899154663086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,4,4,power_law_1.2,0.0686079978942871
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,8,2,balanced,0.02132352113723755
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,8,2,balanced,0.024211199283599855
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,8,2,balanced,0.04166272163391113
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,4,4,power_law_1.2,0.6035545730590821
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,8,2,balanced,0.09459967613220215
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,8,2,balanced,0.12127872467041016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,8,2,balanced,0.1245030403137207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,8,2,balanced,0.1255577564239502
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,8,2,balanced,0.1490828800201416
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,8,2,balanced,0.15003007888793946
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,4,4,power_law_1.2,1.1623999786376953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,8,2,balanced,0.15225855827331541
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,8,2,balanced,0.1521343994140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,8,2,balanced,0.15703167915344238
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,4,4,power_law_1.2,0.43596031188964846
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,8,2,balanced,0.15466752052307128
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,8,2,balanced,0.15736063957214355
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,8,2,balanced,0.1603596878051758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,8,2,balanced,0.1700979232788086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,8,2,balanced,0.17768320083618164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,4,4,power_law_1.2,0.7536563110351563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,4,4,power_law_1.2,1.5395928955078124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,8,2,balanced,0.30628736495971676
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,8,2,balanced,0.4395609664916992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,8,2,balanced,0.476861457824707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,8,2,balanced,0.15563008308410645
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,8,2,balanced,0.720549087524414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,8,2,balanced,0.061999359130859376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,8,2,power_law_1.01,0.0405785608291626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,8,2,power_law_1.01,0.06438528060913086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,8,2,power_law_1.01,0.07955584049224854
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,8,2,balanced,1.2937945556640624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,8,2,power_law_1.01,0.10321536064147949
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,8,2,power_law_1.01,0.10924799919128418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,8,2,power_law_1.01,0.11302271842956543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,8,2,power_law_1.01,0.13775744438171386
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,8,2,power_law_1.01,0.13896063804626463
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,8,2,power_law_1.01,0.023493120670318605
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,8,2,balanced,1.699842529296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,8,2,power_law_1.01,0.1461081600189209
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,8,2,balanced,0.8941043090820312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,8,2,balanced,0.28977920532226564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,8,2,power_law_1.01,0.19024127960205078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,8,2,power_law_1.01,0.2320012855529785
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,8,2,power_law_1.01,0.050378241539001466
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,8,2,power_law_1.01,0.14384127616882325
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,8,2,power_law_1.01,0.14517631530761718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,8,2,power_law_1.01,0.1495577621459961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,8,2,power_law_1.01,0.15903615951538086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,8,2,power_law_1.01,0.16822528839111328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,8,2,power_law_1.01,0.45102462768554685
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,8,2,power_law_1.01,0.5615078353881836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,8,2,power_law_1.01,0.8061695861816407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,8,2,power_law_1.01,1.0032166290283204
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,8,2,power_law_1.01,0.15318016052246092
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,8,2,power_law_1.01,0.284956169128418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,8,2,power_law_1.01,0.3484172821044922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,8,2,power_law_1.2,0.05053823947906494
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,8,2,power_law_1.2,0.062136321067810064
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,8,2,power_law_1.2,0.07157631874084472
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,8,2,power_law_1.2,0.10175488471984864
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,8,2,power_law_1.01,1.4884416198730468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,8,2,power_law_1.2,0.1376460838317871
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,8,2,power_law_1.2,0.13954175949096678
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,8,2,power_law_1.2,0.14329728126525879
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,8,2,power_law_1.2,0.1455667209625244
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,8,2,power_law_1.01,1.9116351318359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,8,2,power_law_1.2,0.14941951751708984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,8,2,power_law_1.2,0.15461376190185547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,8,2,power_law_1.2,0.04011903762817383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,8,2,power_law_1.2,0.15716352462768554
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,8,2,power_law_1.2,0.16407039642333984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,8,2,power_law_1.2,0.17591936111450196
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,8,2,power_law_1.2,0.19839744567871093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,8,2,power_law_1.2,0.23464704513549806
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,8,2,power_law_1.2,0.29138816833496095
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,8,2,power_law_1.2,0.023441920280456545
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,8,2,power_law_1.2,0.3555558395385742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,8,2,power_law_1.2,0.47178112030029296
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,8,2,power_law_1.2,0.5983884811401368
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,8,2,power_law_1.2,0.8176000213623047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,8,2,power_law_1.2,0.10612992286682128
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,8,2,power_law_1.2,0.9959693145751952
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,8,2,power_law_1.2,0.11213439941406249
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,8,2,power_law_1.2,1.5840806579589843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,8,2,power_law_1.2,1.9665830993652342
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,16,1,balanced,0.02834752082824707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,16,1,balanced,0.03757567882537842
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,16,1,balanced,0.05496448040008545
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,16,1,balanced,0.12223936080932618
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,16,1,balanced,0.15720255851745607
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,16,1,balanced,0.15895232200622558
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,16,1,balanced,0.16083648681640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,16,1,balanced,0.16198528289794922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,16,1,balanced,0.16005760192871094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,16,1,balanced,0.07934976100921631
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,16,1,balanced,0.1753753662109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,16,1,balanced,0.16611839294433595
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,16,1,balanced,0.19348543167114257
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,16,1,balanced,0.17786943435668945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,16,1,balanced,0.33408321380615236
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,16,1,balanced,0.18009471893310547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,16,1,balanced,0.17792448043823242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,16,1,balanced,0.4741574478149414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,16,1,balanced,0.504218864440918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,16,1,balanced,0.17940799713134764
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,16,1,power_law_1.01,0.028768000602722166
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,16,1,balanced,0.18225343704223634
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,16,1,balanced,0.7832864379882812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,16,1,balanced,0.18773632049560546
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,16,1,power_law_1.01,0.03892735958099365
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,16,1,balanced,0.9630464172363281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,16,1,power_law_1.01,0.09198592185974122
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,16,1,power_law_1.01,0.1293216037750244
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,16,1,power_law_1.01,0.13475584030151366
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,16,1,balanced,0.32374206542968753
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,16,1,balanced,1.4058143615722656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,16,1,power_law_1.01,0.14740351676940916
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,16,1,power_law_1.01,0.15069184303283692
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,16,1,power_law_1.01,0.16234304428100585
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,16,1,power_law_1.01,0.16535295486450194
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,16,1,power_law_1.01,0.1688800048828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,16,1,power_law_1.01,0.1738003158569336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,16,1,balanced,1.8508773803710938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,16,1,power_law_1.01,0.1747987174987793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,16,1,power_law_1.01,0.18081216812133788
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,16,1,power_law_1.01,0.21889856338500974
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,16,1,power_law_1.01,0.055828480720520025
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,16,1,power_law_1.01,0.24281343460083007
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,16,1,power_law_1.01,0.26501504898071293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,16,1,power_law_1.01,0.30614208221435546
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,16,1,power_law_1.01,0.3779993438720703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,16,1,power_law_1.01,0.140064640045166
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,16,1,power_law_1.01,0.46542015075683596
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,16,1,power_law_1.01,0.5843820953369141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,16,1,power_law_1.2,0.04352447986602783
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,16,1,power_law_1.01,0.8024774169921874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,16,1,power_law_1.2,0.05527232170104981
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,16,1,power_law_1.01,1.0236544036865234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,16,1,power_law_1.2,0.12839360237121583
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,16,1,power_law_1.2,0.14266559600830078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,16,1,power_law_1.01,1.4669049072265625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,16,1,power_law_1.01,0.07370560169219971
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,16,1,power_law_1.2,0.14850624084472658
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,16,1,power_law_1.2,0.028428161144256593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,16,1,power_law_1.2,0.1535859203338623
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,16,1,power_law_1.2,0.16512767791748045
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,16,1,power_law_1.2,0.16867456436157227
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,16,1,power_law_1.2,0.17139455795288086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,16,1,power_law_1.01,1.9251199340820313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,16,1,power_law_1.2,0.18396928787231445
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,16,1,power_law_1.2,0.06988351821899415
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,16,1,power_law_1.2,0.19073919296264646
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,16,1,power_law_1.2,0.08700287818908692
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,16,1,power_law_1.2,0.2136204719543457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,16,1,power_law_1.2,0.22241344451904296
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,16,1,power_law_1.2,0.24802816390991209
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,16,1,power_law_1.2,0.2671769523620605
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,16,1,power_law_1.2,0.31326400756835937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,16,1,power_law_1.2,0.1358291244506836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,16,1,power_law_1.2,0.37927104949951174
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,16,1,power_law_1.2,0.4719526290893555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,1,16,balanced,0.06630911827087402
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,1,16,balanced,0.08854016304016113
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,16,1,power_law_1.2,0.5865280151367187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,1,16,balanced,0.29619455337524414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,1,16,balanced,0.4687500762939453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,1,16,balanced,0.552179183959961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,16,1,power_law_1.2,0.8094828796386718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,1,16,balanced,0.7904064178466796
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,1,16,balanced,0.792906265258789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,1,16,balanced,0.8136627197265625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,1,16,balanced,0.8166310119628907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,1,16,balanced,0.8266175842285156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,16,1,power_law_1.2,1.034185562133789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,1,16,balanced,0.8382208251953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,1,16,balanced,0.1525862407684326
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,1,16,balanced,0.8537522888183593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,1,16,balanced,0.883220443725586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,1,16,balanced,0.8811519622802735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,1,16,balanced,0.90748291015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,1,16,balanced,0.9507942199707031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,1,16,balanced,1.0454054260253907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,1,16,balanced,1.1143283081054687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,1,16,balanced,1.4215206909179687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,1,16,balanced,1.4438873291015626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,16,1,power_law_1.2,1.4804135131835938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,1,16,balanced,0.043589119911193845
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,1,16,balanced,0.056801280975341796
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,1,16,balanced,2.0729190063476564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,1,16,balanced,0.09288960456848144
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,1,16,balanced,0.15456768035888674
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,1,16,balanced,0.2461427116394043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,1,16,balanced,2.469360656738281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,1,16,balanced,0.29388544082641604
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,1,16,balanced,0.3192755126953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,1,16,balanced,0.32129150390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,1,16,balanced,0.33154945373535155
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,1,16,balanced,0.33244800567626953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,1,16,balanced,0.33897727966308594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,1,16,balanced,0.35307777404785157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,1,16,balanced,3.8370764160156248
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,1,16,balanced,0.3768576049804687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,1,16,balanced,0.40076160430908203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,1,16,balanced,0.38655872344970704
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,1,16,balanced,0.4046284866333008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,1,16,balanced,0.43333503723144534
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,1,16,balanced,0.4951129531860351
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,1,16,balanced,0.5625446319580079
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,1,16,balanced,0.6828313446044921
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,1,16,balanced,0.8048268890380859
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,1,16,balanced,1.1514009857177734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,16,1,power_law_1.2,1.9306048583984374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,1,16,balanced,1.386754608154297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,1,16,balanced,4.932866516113281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,1,16,power_law_1.01,0.3763916778564453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,1,16,balanced,2.031130828857422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,1,16,power_law_1.01,0.5239487838745117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,1,16,power_law_1.01,0.49756927490234376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,1,16,balanced,7.456880493164062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,1,16,balanced,2.5270732116699217
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,1,16,power_law_1.01,0.5330598449707031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,1,16,power_law_1.01,0.5403046417236328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,1,16,power_law_1.01,0.4940121459960937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,1,16,power_law_1.01,0.6989823913574218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,1,16,power_law_1.01,0.7694989013671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,1,16,power_law_1.01,0.73259521484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,1,16,power_law_1.01,0.7616268920898437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,1,16,power_law_1.01,0.7446669006347657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,1,16,balanced,3.790571594238281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,1,16,power_law_1.01,0.8013964843749999
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,1,16,power_law_1.01,0.7999616241455078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,1,16,balanced,9.6677001953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,1,16,power_law_1.01,0.8641702270507812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,1,16,power_law_1.01,0.8747827148437499
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,1,16,balanced,4.917369689941406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,1,16,power_law_1.01,0.95378173828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,1,16,power_law_1.01,1.0229209899902343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,1,16,power_law_1.01,1.2000192260742186
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,1,16,power_law_1.01,0.8671231842041015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,1,16,power_law_1.01,1.5125286865234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,1,16,power_law_1.01,1.7419981384277343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,1,16,power_law_1.01,0.19987455368041993
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,1,16,power_law_1.01,2.4234713745117187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,1,16,power_law_1.01,0.27383295059204105
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,1,16,power_law_1.01,0.277702407836914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,1,16,power_law_1.01,0.28226816177368164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,1,16,power_law_1.01,3.135315246582031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,1,16,power_law_1.01,0.31777536392211914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,1,16,power_law_1.01,0.2912819290161133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,1,16,power_law_1.01,0.3141926383972168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,1,16,power_law_1.01,0.3117695999145508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,1,16,power_law_1.01,0.3180723190307617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,1,16,power_law_1.01,4.6943859863281245
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,1,16,power_law_1.01,0.3296819305419922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,1,16,power_law_1.01,0.3426355361938477
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,1,16,power_law_1.01,0.3759833526611328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,1,16,power_law_1.01,0.36184318542480465
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,1,16,power_law_1.01,0.3692659378051758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,1,16,power_law_1.01,0.40790657043457035
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,1,16,power_law_1.01,6.0567169189453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,1,16,power_law_1.01,0.47722496032714845
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,1,16,power_law_1.01,0.5395417785644531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,1,16,power_law_1.01,0.682366714477539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,1,16,power_law_1.01,0.8318463897705077
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,1,16,power_law_1.01,1.1525875091552735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,1,16,power_law_1.01,1.449629364013672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,1,16,power_law_1.01,0.27613439559936526
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,1,16,power_law_1.01,8.909959716796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,1,16,power_law_1.2,0.37717376708984374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,1,16,power_law_1.2,0.4990092849731445
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,1,16,power_law_1.01,2.7684838867187502
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,1,16,power_law_1.2,0.49790206909179685
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,1,16,power_law_1.01,2.0111859130859377
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,1,16,power_law_1.01,0.27069055557250976
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,1,16,power_law_1.2,0.5346598434448242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,1,16,power_law_1.2,0.5288435363769531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,1,16,power_law_1.2,0.5338572692871094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,1,16,power_law_1.2,0.7511219024658203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,1,16,power_law_1.2,0.7064320373535156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,1,16,power_law_1.2,0.7099954986572266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,1,16,power_law_1.01,4.1851263427734375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,1,16,power_law_1.2,0.7243405151367187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,1,16,power_law_1.2,0.7650547027587891
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,1,16,power_law_1.01,12.703057861328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,1,16,power_law_1.2,0.8039910125732422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,1,16,power_law_1.2,0.8393791961669923
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,1,16,power_law_1.2,0.8890290832519531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,1,16,power_law_1.2,0.8787955474853515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,1,16,power_law_1.2,0.9375437164306641
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,1,16,power_law_1.01,5.373153076171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,1,16,power_law_1.2,1.0081088256835937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,1,16,power_law_1.2,1.0801689910888672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,1,16,power_law_1.2,1.2565824127197265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,1,16,power_law_1.2,1.6446450805664063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,1,16,power_law_1.2,1.9458163452148436
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,1,16,power_law_1.2,0.1996633529663086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,1,16,power_law_1.2,0.2573311996459961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,1,16,power_law_1.2,0.2755353546142578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,1,16,power_law_1.2,2.898401184082031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,1,16,power_law_1.2,0.28034048080444335
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,1,16,power_law_1.2,0.28389888763427734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,1,16,power_law_1.2,0.24945663452148437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,1,16,power_law_1.2,0.3166592025756836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,1,16,power_law_1.2,0.30337919235229494
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,1,16,power_law_1.2,0.3077952003479004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,1,16,power_law_1.2,3.8749349975585936
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,1,16,power_law_1.2,0.3126873588562012
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,1,16,power_law_1.2,0.3200076675415039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,1,16,power_law_1.2,0.3275532913208008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,1,16,power_law_1.2,0.34349056243896486
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,1,16,power_law_1.2,0.37456512451171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,1,16,power_law_1.2,0.3680819320678711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,1,16,power_law_1.2,0.3909107208251953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,1,16,power_law_1.2,0.435203857421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,1,16,power_law_1.2,5.197113647460937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,1,16,power_law_1.2,0.5052441787719727
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,1,16,power_law_1.2,0.5631167984008789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,1,16,power_law_1.2,0.7331251525878907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,1,16,power_law_1.2,0.9124237060546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,1,16,power_law_1.2,1.2803135681152344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,1,16,power_law_1.2,1.6560435485839844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,1,16,power_law_1.2,7.525151977539062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,2,8,balanced,0.05564288139343262
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,2,8,balanced,0.12638336181640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,2,8,balanced,0.08377984046936035
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,2,8,balanced,0.25593215942382813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,2,8,balanced,0.40336639404296876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,2,8,balanced,0.5401279830932617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,2,8,balanced,0.5496960067749024
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,1,16,power_law_1.2,2.3673114013671874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,2,8,balanced,0.5582182312011719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,1,16,power_law_1.2,3.1749887084960937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,2,8,balanced,0.7800511932373047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,2,8,balanced,0.7963699340820313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,2,8,balanced,0.7828377532958984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,2,8,balanced,0.8207794952392579
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,2,8,balanced,0.8052710723876952
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,2,8,balanced,0.8496934509277343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,2,8,balanced,0.8481062316894532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,2,8,balanced,0.8710310363769531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,2,8,balanced,0.9124441528320313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,2,8,balanced,1.0026470184326173
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,2,8,balanced,1.0649113464355469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,2,8,balanced,1.2384012603759766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,1,16,power_law_1.2,4.561863708496094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,2,8,balanced,1.3706227111816407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,1,16,power_law_1.2,11.6203515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,2,8,balanced,1.9320614624023438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,2,8,balanced,0.03670144081115723
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,2,8,balanced,2.2402163696289064
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,1,16,power_law_1.2,15.295772705078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,2,8,balanced,0.15455360412597657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,2,8,balanced,0.215548152923584
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,2,8,balanced,0.05209343910217286
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,2,8,balanced,0.28534400939941407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,2,8,balanced,3.392569580078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,2,8,balanced,0.3004991912841797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,2,8,balanced,0.30502656936645506
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,2,8,balanced,0.33882495880126956
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,2,8,balanced,0.34218238830566405
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,2,8,balanced,0.3449663925170898
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,2,8,balanced,0.3616960144042969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,2,8,balanced,4.337387390136719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,2,8,balanced,0.37795841217041015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,2,8,balanced,0.3984102249145508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,2,8,balanced,0.3833740615844726
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,2,8,balanced,0.3965862274169922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,2,8,balanced,0.4222079849243164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,2,8,balanced,0.4665740966796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,2,8,balanced,0.5164659118652344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,2,8,balanced,0.6204172897338867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,1,16,power_law_1.2,5.955303955078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,2,8,balanced,0.08699775695800781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,2,8,balanced,0.7088230133056641
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,2,8,balanced,6.65014404296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,2,8,balanced,1.0260723114013672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,2,8,balanced,1.2268326568603516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,2,8,power_law_1.01,0.2310028839111328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,2,8,power_law_1.01,0.2923417663574219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,2,8,power_law_1.01,0.3705075073242187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,2,8,balanced,1.7702656555175782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,2,8,power_law_1.01,0.3873023986816406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,2,8,balanced,2.1606130981445313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,2,8,power_law_1.01,0.4959513473510742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,2,8,power_law_1.01,0.4855795288085938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,2,8,power_law_1.01,0.4559756851196289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,2,8,balanced,8.641025390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,2,8,power_law_1.01,0.48291198730468754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,2,8,power_law_1.01,0.6744755554199219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,2,8,balanced,3.2604736328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,2,8,power_law_1.01,0.7354163360595704
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,2,8,power_law_1.01,0.7242444610595703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,2,8,power_law_1.01,0.7617625427246094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,2,8,power_law_1.01,0.8235724639892578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,2,8,power_law_1.01,0.8196351623535156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,2,8,power_law_1.01,0.8455513763427736
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,2,8,balanced,4.169546203613281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,2,8,power_law_1.01,1.0203033447265626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,2,8,power_law_1.01,1.0760934448242188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,2,8,power_law_1.01,0.7732454681396485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,2,8,power_law_1.01,1.331741485595703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,2,8,power_law_1.01,0.8901324462890624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,2,8,power_law_1.01,1.5717388916015627
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,2,8,power_law_1.01,0.11419391632080078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,2,8,power_law_1.01,0.20244224548339843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,2,8,power_law_1.01,2.1062886047363283
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,2,8,power_law_1.01,0.17216768264770507
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,2,8,power_law_1.01,0.21321088790893555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,2,8,power_law_1.01,0.2575014305114746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,2,8,power_law_1.01,0.26592895507812503
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,2,8,power_law_1.01,2.6808331298828127
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,2,8,power_law_1.01,0.27023103713989255
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,2,8,power_law_1.01,3.878050537109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,2,8,power_law_1.01,4.916611938476562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,2,8,power_law_1.01,0.2742950439453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,2,8,power_law_1.01,7.050748291015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,2,8,power_law_1.01,0.3584115219116211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,2,8,power_law_1.01,0.375467529296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,2,8,power_law_1.01,0.3056230354309082
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,2,8,power_law_1.01,0.4066022491455078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,2,8,power_law_1.01,0.3172710418701172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,2,8,power_law_1.01,0.31544319152832034
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,2,8,power_law_1.01,9.41747314453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,2,8,power_law_1.01,0.4676595306396484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,2,8,power_law_1.01,0.5285926437377929
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,2,8,power_law_1.01,0.6519129943847657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,2,8,power_law_1.01,0.32784385681152345
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,2,8,power_law_1.01,0.7727910614013671
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,2,8,power_law_1.01,1.2871347045898438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,2,8,power_law_1.01,0.34234497070312503
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,2,8,power_law_1.2,0.2315443229675293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,2,8,power_law_1.01,1.8208370971679688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,2,8,power_law_1.2,0.2781017684936523
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,2,8,power_law_1.2,0.2915007972717285
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,2,8,power_law_1.01,2.3595674133300784
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,2,8,power_law_1.2,0.3787251281738281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,2,8,power_law_1.2,0.4529919815063477
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,2,8,power_law_1.2,0.5117567825317383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,2,8,power_law_1.01,0.3797529602050781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,2,8,power_law_1.2,0.49468929290771485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,2,8,power_law_1.2,0.6795327758789063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,2,8,power_law_1.01,3.4388940429687502
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,2,8,power_law_1.2,0.6902028656005859
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,2,8,power_law_1.2,0.7395442962646485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,2,8,power_law_1.2,0.7808882904052734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,2,8,power_law_1.2,0.44634750366210935
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,2,8,power_law_1.2,0.8054195404052734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,2,8,power_law_1.2,0.842525405883789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,2,8,power_law_1.01,4.569605102539063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,2,8,power_law_1.2,0.8768946838378906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,2,8,power_law_1.2,0.9499712371826172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,2,8,power_law_1.2,1.0163801574707032
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,2,8,power_law_1.2,1.1436557006835937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,2,8,power_law_1.2,0.8397542572021484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,2,8,power_law_1.2,1.423354949951172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,2,8,power_law_1.01,1.02110595703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,2,8,power_law_1.2,1.6785459899902342
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,2,8,power_law_1.2,0.11483263969421387
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,2,8,power_law_1.2,0.1566592025756836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,2,8,power_law_1.2,2.23836669921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,2,8,power_law_1.2,0.17786239624023437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,2,8,power_law_1.2,0.20353664398193358
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,2,8,power_law_1.2,0.24407167434692384
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,2,8,power_law_1.2,2.8700967407226563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,2,8,power_law_1.2,0.2683033561706543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,2,8,power_law_1.2,0.27153663635253905
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,2,8,power_law_1.2,0.26938495635986326
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,2,8,power_law_1.2,0.29372928619384764
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,2,8,power_law_1.2,0.3065318489074707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,2,8,power_law_1.2,0.314967041015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,2,8,power_law_1.2,0.33631744384765627
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,2,8,power_law_1.2,0.3599795150756836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,2,8,power_law_1.2,0.39372161865234373
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,2,8,power_law_1.2,0.3770444869995117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,2,8,power_law_1.2,4.1470617675781245
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,2,8,power_law_1.2,0.3902463912963867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,2,8,power_law_1.2,0.4230783843994141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,2,8,power_law_1.2,0.48680065155029295
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,2,8,power_law_1.2,0.5417062377929687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,2,8,power_law_1.2,0.6681779479980469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,2,8,power_law_1.2,5.388704223632812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,2,8,power_law_1.2,0.8082431793212891
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,2,8,power_law_1.2,1.056473617553711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,2,8,power_law_1.2,1.3425765991210938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,4,4,balanced,0.05306367874145508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,4,4,balanced,0.07758975982666015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,4,4,balanced,0.11737343788146973
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,4,4,balanced,0.24311424255371095
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,2,8,power_law_1.2,1.9376307678222655
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,4,4,balanced,0.3939750289916992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,4,4,balanced,0.5411967849731445
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,4,4,balanced,0.5545676803588867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,4,4,balanced,0.5578073501586914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,2,8,power_law_1.2,2.511962890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,4,4,balanced,0.5704793548583984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,4,4,balanced,0.5737919998168945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,4,4,balanced,0.5829158401489257
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,4,4,balanced,0.8107263946533203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,2,8,power_law_1.2,7.702566528320313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,4,4,balanced,0.8146150207519531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,4,4,balanced,0.8445708465576172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,4,4,balanced,0.8383692932128906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,4,4,balanced,0.8563648223876953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,4,4,balanced,0.8935897827148438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,4,4,balanced,0.9715494537353516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,4,4,balanced,1.0299404907226564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,2,8,power_law_1.2,10.50688720703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,4,4,balanced,1.2152217864990233
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,4,4,balanced,1.335900115966797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,4,4,balanced,1.8570509338378904
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,4,4,balanced,0.03930367946624756
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,4,4,balanced,2.1486245727539064
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,2,8,power_law_1.2,3.6620913696289064
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,4,4,balanced,0.050919680595397955
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,4,4,balanced,0.07719552040100097
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,4,4,balanced,0.13666048049926757
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,4,4,balanced,0.21477632522583007
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,4,4,balanced,0.292938232421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,4,4,balanced,0.30508031845092776
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,4,4,balanced,0.3087385559082031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,4,4,balanced,3.2401599121093754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,4,4,balanced,0.3190092849731445
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,4,4,balanced,0.3236748886108399
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,4,4,balanced,0.3305728149414063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,4,4,balanced,0.38540672302246093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,4,4,balanced,0.39356800079345705
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,4,4,balanced,0.4222758483886719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,4,4,balanced,4.0673675537109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,4,4,balanced,0.40018177032470703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,4,4,balanced,0.4072563171386719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,4,4,balanced,0.4410611343383789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,4,4,balanced,0.48395648956298826
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,4,4,balanced,0.5235635375976562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,4,4,balanced,0.6070297622680665
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,4,4,balanced,0.6919296264648438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,4,4,balanced,1.034664993286133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,4,4,balanced,1.1968819427490236
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,4,4,balanced,6.0182733154296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,4,4,power_law_1.01,0.10849535942077637
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,2,8,power_law_1.2,4.828574829101562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,4,4,balanced,1.7416972351074218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,4,4,power_law_1.01,0.235361270904541
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,4,4,power_law_1.01,0.20351743698120117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,4,4,power_law_1.01,0.271333122253418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,4,4,balanced,2.077250518798828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,4,4,power_law_1.01,0.32700801849365235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,4,4,power_law_1.01,0.48184833526611326
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,4,4,power_law_1.01,0.4463065719604492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,4,4,power_law_1.01,0.5153087997436524
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,4,4,power_law_1.01,0.4783744049072266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,4,4,power_law_1.01,0.5224409484863282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,4,4,balanced,3.175486755371094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,4,4,power_law_1.01,0.7323865509033203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,4,4,power_law_1.01,0.46828670501708985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,4,4,power_law_1.01,0.7494310760498047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,4,4,power_law_1.01,0.7918476867675781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,4,4,power_law_1.01,0.8050713348388673
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,4,4,balanced,7.7546276855468745
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,4,4,power_law_1.01,0.8647052764892578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,4,4,balanced,4.022593383789062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,4,4,power_law_1.01,0.9688384246826172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,4,4,power_law_1.01,1.025396499633789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,4,4,power_law_1.01,1.218513946533203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,4,4,power_law_1.01,1.4365005493164062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,4,4,power_law_1.01,0.817053451538086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,4,4,power_law_1.01,0.07058176040649414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,4,4,power_law_1.01,0.11727104187011719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,4,4,power_law_1.01,1.906439666748047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,4,4,power_law_1.01,0.1162559986114502
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,4,4,power_law_1.01,0.15266176223754885
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,4,4,power_law_1.01,2.4394406127929686
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,4,4,power_law_1.01,0.19358976364135744
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,4,4,power_law_1.01,0.24494592666625978
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,4,4,power_law_1.01,0.274330883026123
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,4,4,power_law_1.01,0.275097599029541
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,4,4,power_law_1.01,0.29328895568847657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,4,4,power_law_1.01,0.29458944320678715
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,4,4,power_law_1.01,3.44898681640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,4,4,power_law_1.01,0.36296192169189456
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,4,4,power_law_1.01,0.37436542510986326
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,4,4,power_law_1.01,0.39440383911132815
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,4,4,power_law_1.01,0.3825279998779297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,4,4,power_law_1.01,0.3926067352294922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,4,4,power_law_1.01,4.426976013183594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,4,4,power_law_1.01,0.4256409454345703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,4,4,power_law_1.01,0.4916479873657226
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,4,4,power_law_1.01,0.5345024108886719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,4,4,power_law_1.01,0.6516544342041015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,4,4,power_law_1.01,0.7695283508300781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,4,4,power_law_1.01,1.0056896209716797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,4,4,power_law_1.01,6.337300415039063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,4,4,power_law_1.01,1.263289566040039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,4,4,power_law_1.2,0.11901056289672851
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,4,4,power_law_1.2,0.1551155185699463
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,4,4,power_law_1.01,0.3015936088562012
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,4,4,power_law_1.2,0.20552576065063474
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,4,4,power_law_1.2,0.2714752006530762
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,4,4,power_law_1.01,2.285212097167969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,4,4,power_law_1.2,0.3061427116394043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,4,4,power_law_1.2,0.4716172790527343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,4,4,power_law_1.2,0.4284236907958984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,4,4,power_law_1.01,1.7812121582031248
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,4,4,power_law_1.2,0.4814656066894531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,4,4,power_law_1.01,8.503424072265625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,4,4,power_law_1.2,0.4879859161376953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,4,4,power_law_1.2,0.5349004745483399
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,4,4,power_law_1.01,3.346152954101562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,4,4,power_law_1.2,0.7755072021484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,4,4,power_law_1.2,0.7825459289550781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,4,4,power_law_1.2,0.8215090942382812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,4,4,power_law_1.2,0.8242009735107422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,4,4,power_law_1.2,0.8551462554931641
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,4,4,power_law_1.2,0.9037964630126953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,4,4,power_law_1.2,0.9881062316894532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,4,4,power_law_1.2,1.0749260711669923
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,4,4,power_law_1.2,1.2653427124023438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,4,4,power_law_1.2,1.482787780761719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,4,4,power_law_1.2,0.4736896133422851
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,4,4,power_law_1.01,4.318809509277344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,4,4,power_law_1.2,0.07015552043914794
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,4,4,power_law_1.2,1.9866648864746093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,4,4,power_law_1.2,0.10650624275207518
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,4,4,power_law_1.2,0.11812352180480956
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,4,4,power_law_1.2,0.15612544059753417
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,4,4,power_law_1.2,0.16536832809448243
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,4,4,power_law_1.2,2.572833251953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,4,4,power_law_1.2,0.24580991744995115
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,4,4,power_law_1.2,0.26829952239990235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,4,4,power_law_1.2,0.2722112083435059
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,4,4,power_law_1.2,0.29342592239379883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,4,4,power_law_1.2,0.3010534477233887
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,4,4,power_law_1.2,3.4773031616210934
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,4,4,power_law_1.2,0.3655321502685547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,4,4,power_law_1.2,0.37932544708251953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,4,4,power_law_1.2,0.4103936004638672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,4,4,power_law_1.2,0.3975884628295899
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,4,4,power_law_1.2,0.4036057662963867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,4,4,power_law_1.2,0.4554572677612304
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,4,4,power_law_1.2,0.4936217498779297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,4,4,power_law_1.2,4.60076416015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,4,4,power_law_1.2,0.555654411315918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,4,4,power_law_1.2,0.6696422576904297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,4,4,power_law_1.2,0.7950067138671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,4,4,power_law_1.2,1.0270809936523437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,4,4,power_law_1.2,1.3358963012695313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,8,2,balanced,0.05181951999664307
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,8,2,balanced,0.07091072082519531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,4,4,power_law_1.2,1.8340377807617188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,8,2,balanced,0.1130508804321289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,8,2,balanced,0.2449139213562012
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,8,2,balanced,0.39587711334228515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,8,2,balanced,0.5549798583984374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,4,4,power_law_1.2,0.2720716857910156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,8,2,balanced,0.5666355133056641
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,4,4,power_law_1.2,6.641534423828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,4,4,power_law_1.2,2.3818585205078127
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,8,2,balanced,0.5697932815551757
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,8,2,balanced,0.5787776184082032
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,8,2,balanced,0.5879769515991211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,8,2,balanced,0.5985702514648438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,8,2,balanced,0.6126515197753906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,8,2,balanced,0.6193382263183593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,8,2,balanced,0.6396236801147461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,4,4,power_law_1.2,9.43631103515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,8,2,balanced,0.8216690826416017
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,8,2,balanced,0.8339852905273437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,8,2,balanced,0.8660095977783204
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,8,2,balanced,0.9341248321533204
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,8,2,balanced,1.0013286590576171
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,8,2,balanced,1.2944525146484376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,4,4,power_law_1.2,3.429165954589844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,8,2,balanced,1.4155967712402344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,8,2,balanced,1.9653567504882812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,8,2,balanced,0.03777663946151734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,8,2,balanced,2.2503257751464845
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,8,2,balanced,0.05194367885589599
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,8,2,balanced,0.08659199714660645
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,8,2,balanced,0.14376192092895507
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,8,2,balanced,0.23921791076660157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,8,2,balanced,0.33610622406005863
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,8,2,balanced,0.3455974578857422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,8,2,balanced,3.4386520385742188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,8,2,balanced,0.3485222244262695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,8,2,balanced,0.35711105346679684
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,8,2,balanced,0.36111103057861327
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,8,2,balanced,0.3662169647216797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,8,2,balanced,0.37790721893310547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,8,2,balanced,4.317550048828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,8,2,balanced,0.4023551940917969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,8,2,balanced,0.44605567932128903
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,8,2,balanced,0.4588544082641602
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,8,2,balanced,0.46965503692626953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,8,2,balanced,0.5138137435913086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,8,2,balanced,0.5481228637695312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,8,2,balanced,0.3828607940673828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,8,2,balanced,0.6232601547241211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,8,2,balanced,0.6997337341308594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,8,2,balanced,1.1043827056884765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,4,4,power_law_1.2,4.398766174316406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,8,2,balanced,6.32972900390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,8,2,balanced,1.2589900970458985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,8,2,power_law_1.01,0.0751743984222412
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,8,2,power_law_1.01,0.11721216201782228
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,8,2,power_law_1.01,0.14055423736572265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,8,2,balanced,1.8374592590332033
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,8,2,power_law_1.01,0.24797439575195312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,8,2,power_law_1.01,0.30833408355712894
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,8,2,balanced,2.133251800537109
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,8,2,power_law_1.01,0.4537996673583985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,8,2,power_law_1.01,0.4698880004882812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,8,2,power_law_1.01,0.47635326385498045
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,8,2,power_law_1.01,0.490777587890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,8,2,power_law_1.01,0.512334098815918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,8,2,power_law_1.01,0.5496409606933594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,8,2,power_law_1.01,0.5794483184814453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,8,2,power_law_1.01,0.5931814575195313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,8,2,balanced,3.299059143066406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,8,2,power_law_1.01,0.6657279968261719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,8,2,power_law_1.01,0.7706956481933593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,8,2,power_law_1.01,0.786376953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,8,2,balanced,8.115558471679687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,8,2,power_law_1.01,0.8370738983154297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,8,2,power_law_1.01,0.9478105926513672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,8,2,power_law_1.01,1.0187494659423828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,8,2,power_law_1.01,1.2198809814453124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,8,2,power_law_1.01,1.4326016235351564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,8,2,power_law_1.01,0.04942463874816895
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,8,2,power_law_1.01,1.8992767333984375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,8,2,balanced,4.111512451171874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,8,2,power_law_1.01,0.10122367858886719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,8,2,power_law_1.01,0.07534336090087891
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,8,2,power_law_1.01,2.407124481201172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,8,2,power_law_1.01,0.1492633628845215
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,8,2,power_law_1.01,0.18094463348388673
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,8,2,power_law_1.01,0.2764505577087403
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,8,2,power_law_1.01,0.29373567581176757
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,8,2,power_law_1.01,0.3013260841369629
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,8,2,power_law_1.01,0.32028545379638673
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,8,2,power_law_1.01,0.3235622406005859
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,8,2,power_law_1.01,3.381410522460938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,8,2,power_law_1.01,0.33323265075683595
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,8,2,power_law_1.01,0.3437081527709961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,8,2,power_law_1.01,0.3579635238647461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,8,2,power_law_1.01,0.37675521850585936
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,8,2,power_law_1.01,0.42344577789306637
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,8,2,power_law_1.01,0.43344127655029296
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,8,2,power_law_1.01,0.44872192382812504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,8,2,power_law_1.01,4.441254272460937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,8,2,power_law_1.01,0.5241766357421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,8,2,power_law_1.01,0.5734195327758789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,8,2,power_law_1.01,0.6906368255615234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,8,2,power_law_1.01,0.808321304321289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,8,2,power_law_1.01,1.0376947021484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,8,2,power_law_1.2,0.0687014389038086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,8,2,power_law_1.01,1.3113395690917968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,8,2,power_law_1.2,0.09659903526306153
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,8,2,power_law_1.2,0.1423692798614502
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,8,2,power_law_1.01,1.788907470703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,8,2,power_law_1.2,0.24000127792358397
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,8,2,power_law_1.2,0.2848512077331543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,8,2,power_law_1.01,2.294151611328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,8,2,power_law_1.2,0.442237434387207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,8,2,power_law_1.2,0.45991039276123047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,8,2,power_law_1.2,0.4705868911743164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,8,2,power_law_1.01,6.375740356445313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,8,2,power_law_1.2,0.49869823455810547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,8,2,power_law_1.2,0.5115955352783204
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,8,2,power_law_1.2,0.5546265411376953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,8,2,power_law_1.01,3.36334716796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,8,2,power_law_1.2,0.5794956970214844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,8,2,power_law_1.01,8.403655395507812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,8,2,power_law_1.2,0.6030118560791016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,8,2,power_law_1.2,0.6815180969238281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,8,2,power_law_1.2,0.7984268951416016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,8,2,power_law_1.2,0.8227737426757813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,8,2,power_law_1.2,0.8646272277832031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,8,2,power_law_1.01,4.333926391601563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,8,2,power_law_1.2,0.947546844482422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,8,2,power_law_1.2,1.046091537475586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,8,2,power_law_1.2,1.2520025634765626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,8,2,power_law_1.2,1.464111328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,8,2,power_law_1.2,0.04965375900268555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,8,2,power_law_1.2,1.929265899658203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,8,2,power_law_1.2,0.06954368114471435
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,8,2,power_law_1.2,0.10011775970458985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,8,2,power_law_1.2,0.1314086437225342
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,8,2,power_law_1.2,2.505675506591797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,8,2,power_law_1.2,0.1726553535461426
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,8,2,power_law_1.2,0.27494144439697266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,8,2,power_law_1.2,0.28976640701293943
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,8,2,power_law_1.2,0.29895679473876957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,8,2,power_law_1.2,0.31025407791137694
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,8,2,power_law_1.2,3.4625048828125005
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,8,2,power_law_1.2,0.3237401580810547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,8,2,power_law_1.2,0.3384435272216797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,8,2,power_law_1.2,0.35387649536132815
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,8,2,power_law_1.2,0.3620377731323242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,8,2,power_law_1.2,0.3905279922485351
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,8,2,power_law_1.2,0.4353830337524414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,8,2,power_law_1.2,0.4494451141357422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,8,2,power_law_1.2,0.4790771102905273
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,8,2,power_law_1.2,4.574953002929687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,8,2,power_law_1.2,0.5396531295776368
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,8,2,power_law_1.2,0.5929318237304687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,8,2,power_law_1.2,0.7074687957763672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,8,2,power_law_1.2,0.8224063873291015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,8,2,power_law_1.2,1.0640076446533204
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,8,2,power_law_1.2,1.3508543395996093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,8,2,power_law_1.2,1.8222399902343749
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,8,2,power_law_1.2,2.3625701904296874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,8,2,power_law_1.2,6.480587768554687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,16,1,balanced,0.08930432319641113
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,16,1,balanced,0.1276473617553711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,8,2,power_law_1.2,8.755195922851563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,16,1,balanced,0.20710336685180666
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,16,1,balanced,0.4221747207641601
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,8,2,power_law_1.2,3.4232974243164067
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,16,1,balanced,0.73621826171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,16,1,balanced,1.050774383544922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,16,1,balanced,1.0615430450439454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,16,1,balanced,1.0733395385742188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,16,1,balanced,1.0816134643554687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,16,1,balanced,1.1253459167480468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,16,1,balanced,1.1519737243652344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,16,1,balanced,1.1491353607177734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,16,1,balanced,1.0898233795166017
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,16,1,balanced,1.0967628479003906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,16,1,balanced,1.4869030761718751
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,16,1,balanced,1.8962200927734376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,8,2,power_law_1.2,4.289612731933594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,16,1,balanced,1.1200179290771484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,16,1,balanced,1.4527218627929688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,16,1,balanced,1.5252607727050782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,16,1,balanced,2.372344970703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,16,1,balanced,2.5025868225097656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,16,1,balanced,3.5372216796875002
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,16,1,balanced,3.797132263183594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,16,1,balanced,5.972644653320312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,16,1,balanced,7.316470336914063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,16,1,balanced,10.713746337890624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,16,1,balanced,0.5229945755004882
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,16,1,balanced,0.9733990478515626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,16,1,balanced,13.356596679687499
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,16,1,balanced,1.8397331237792969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,16,1,balanced,3.530078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,16,1,balanced,6.970851440429688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,16,1,balanced,10.181162109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,16,1,balanced,10.171166381835937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,16,1,balanced,10.17878662109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,16,1,balanced,10.205535278320312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,16,1,balanced,10.219631958007813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,16,1,balanced,10.257991943359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,16,1,balanced,10.249779052734375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,16,1,balanced,10.135344848632812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,16,1,balanced,10.238748168945312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,16,1,balanced,10.261536865234374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,16,1,balanced,10.305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,16,1,balanced,10.309185791015626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,16,1,balanced,8.680220947265624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,16,1,balanced,8.760503540039062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,16,1,balanced,9.066792602539063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,16,1,balanced,9.046887817382814
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,16,1,power_law_1.01,0.08854975700378417
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,16,1,power_law_1.01,0.2081612777709961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,16,1,power_law_1.01,0.12662015914916994
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,16,1,power_law_1.01,0.502999038696289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,16,1,power_law_1.01,0.3803641510009766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,16,1,power_law_1.01,0.7972640228271485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,16,1,power_law_1.01,0.8502899169921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,16,1,power_law_1.01,0.8832640075683594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,16,1,power_law_1.01,0.9161689758300782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,16,1,power_law_1.01,0.951190414428711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,16,1,power_law_1.01,1.024676513671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,16,1,power_law_1.01,1.0898585510253906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,16,1,power_law_1.01,1.1485887908935548
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,16,1,balanced,17.478612060546872
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,16,1,power_law_1.01,1.3066624450683595
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,16,1,power_law_1.01,1.4087954711914061
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,16,1,power_law_1.01,1.5622476196289061
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,16,1,power_law_1.01,1.8587379455566406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,16,1,power_law_1.01,1.56374267578125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,16,1,balanced,17.704000244140623
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,16,1,power_law_1.01,1.7311929321289063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,16,1,power_law_1.01,2.078059539794922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,16,1,power_law_1.01,2.4473785400390624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,16,1,power_law_1.01,3.212165222167969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,16,1,power_law_1.01,3.9966278076171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,16,1,power_law_1.01,5.591580200195312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,16,1,balanced,26.24220703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,16,1,power_law_1.01,7.172812500000001
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,16,1,balanced,27.1849609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,16,1,power_law_1.01,10.37727783203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,16,1,power_law_1.01,0.5381171035766601
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,16,1,power_law_1.01,13.56064453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,16,1,power_law_1.01,0.9851475524902344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,16,1,power_law_1.01,1.8321618652343752
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,16,1,power_law_1.01,3.2146746826171877
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,16,1,balanced,43.6434326171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,16,1,power_law_1.01,7.609254150390624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,16,1,power_law_1.01,8.007509765624999
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,16,1,power_law_1.01,8.3472119140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,16,1,power_law_1.01,4.5536062622070315
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,16,1,power_law_1.01,8.493275146484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,16,1,power_law_1.01,8.701763916015626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,16,1,power_law_1.01,8.880128784179687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,16,1,power_law_1.01,9.125885620117188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,16,1,balanced,53.2296533203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,16,1,power_law_1.01,9.093601684570313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,16,1,power_law_1.01,9.350679931640624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,16,1,power_law_1.01,9.649593505859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,16,1,power_law_1.01,10.055700073242187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,16,1,power_law_1.01,8.805048828124999
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,16,1,power_law_1.01,10.85408447265625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,16,1,power_law_1.01,9.571142578125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,16,1,power_law_1.01,10.84346923828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,16,1,power_law_1.2,0.08881024360656739
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,16,1,power_law_1.2,0.12713919639587404
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,16,1,power_law_1.2,0.20820991516113282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,16,1,power_law_1.01,12.4332080078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,16,1,power_law_1.2,0.3226412963867188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,16,1,power_law_1.2,0.46506111145019535
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,16,1,power_law_1.2,0.7819225311279296
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,16,1,power_law_1.2,0.8454093170166015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,16,1,power_law_1.2,0.8717183685302734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,16,1,power_law_1.2,0.9200383758544921
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,16,1,power_law_1.01,15.432559814453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,16,1,power_law_1.2,0.9849081420898438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,16,1,power_law_1.2,1.0443276977539062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,16,1,power_law_1.2,1.20212158203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,16,1,power_law_1.2,1.3393331909179689
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,16,1,power_law_1.2,1.1174329376220702
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,16,1,power_law_1.01,18.230941162109374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,16,1,power_law_1.2,1.4653305053710937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,16,1,power_law_1.2,1.6053485107421874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,16,1,power_law_1.2,1.6245989990234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,16,1,power_law_1.2,1.8886009216308595
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,16,1,power_law_1.2,1.7986553955078126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,16,1,power_law_1.2,2.1305247497558595
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,16,1,power_law_1.2,2.485125732421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,16,1,power_law_1.2,3.2600946044921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,16,1,power_law_1.2,4.027939758300781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,16,1,power_law_1.01,24.15614990234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,16,1,power_law_1.2,0.5262732696533203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,16,1,power_law_1.2,5.619280395507813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,16,1,power_law_1.2,0.9693984222412109
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,16,1,power_law_1.2,2.9231756591796874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,16,1,power_law_1.2,7.200057373046874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,16,1,power_law_1.01,30.12156982421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,16,1,power_law_1.2,4.018899230957031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,16,1,power_law_1.2,7.376395874023437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,16,1,power_law_1.2,10.373374023437501
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,16,1,power_law_1.2,1.8398374938964843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,16,1,power_law_1.2,8.028515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,16,1,power_law_1.2,8.165077514648438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,16,1,power_law_1.01,42.8085302734375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,16,1,power_law_1.2,8.515479736328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,16,1,power_law_1.2,8.791683349609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,16,1,power_law_1.2,9.043259887695312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,16,1,power_law_1.2,9.24420166015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,16,1,power_law_1.2,9.259745483398437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,16,1,power_law_1.2,13.553183593750001
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,16,1,power_law_1.2,9.943750610351563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,16,1,power_law_1.2,10.31779296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,16,1,power_law_1.2,10.665924072265625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,16,1,power_law_1.01,55.107026367187494
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,16,1,power_law_1.2,9.359699096679687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,16,1,power_law_1.2,11.453226318359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,16,1,power_law_1.2,10.043034667968751
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,16,1,power_law_1.2,11.438955078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,4,8,balanced,0.06198400020599365
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,4,8,balanced,0.06200448036193847
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,4,8,balanced,0.0614463996887207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,4,8,balanced,0.1089305591583252
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,4,8,balanced,0.1096243190765381
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,4,8,balanced,0.10949503898620605
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,4,8,balanced,0.11036288261413574
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,4,8,balanced,0.11067008018493651
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,4,8,balanced,0.11042431831359863
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,16,1,power_law_1.2,12.965150146484376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,4,8,balanced,0.111144962310791
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,4,8,balanced,0.11127936363220214
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,4,8,balanced,0.11513471603393555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,4,8,balanced,0.11433471679687499
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,4,8,balanced,0.11778176307678223
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,4,8,balanced,0.11776384353637695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,4,8,balanced,0.12309375762939453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,4,8,balanced,0.12630528450012207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,4,8,balanced,0.11426943778991698
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,4,8,balanced,0.13102080345153808
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,4,8,balanced,0.17950719833374024
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,4,8,balanced,0.19817472457885743
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,4,8,balanced,0.3134041595458984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,4,8,balanced,0.3754483032226562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,4,8,balanced,0.5331430435180664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,4,8,balanced,0.7070118713378906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,4,8,balanced,1.0306534576416015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,4,8,balanced,0.03614207983016968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,4,8,balanced,0.036366078853607174
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,4,8,balanced,0.036081919670104975
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,4,8,balanced,1.3816487121582033
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,4,8,balanced,0.05729536056518555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,4,8,balanced,0.05997183799743653
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,4,8,balanced,0.05783679962158204
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,16,1,power_law_1.2,15.827513427734374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,4,8,balanced,0.05962751865386963
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,4,8,balanced,0.0574451208114624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,4,8,balanced,0.06007423877716065
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,4,8,balanced,0.05800191879272461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,4,8,balanced,0.05993087768554688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,4,8,balanced,0.05874815940856933
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,4,8,balanced,0.06122623920440674
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,4,8,balanced,0.06049280166625977
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,4,8,balanced,0.06256768226623535
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,4,8,balanced,0.06449535846710205
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,4,8,balanced,0.06914048194885254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,4,8,balanced,0.07095424175262452
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,4,8,balanced,0.05803391933441162
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,4,8,balanced,0.07744256019592285
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,4,8,balanced,0.10748288154602051
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,4,8,balanced,0.12392704010009765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,4,8,balanced,0.17717248916625977
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,4,8,balanced,0.22336000442504883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,4,8,balanced,0.3274201583862305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,4,8,balanced,0.4119295883178711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,4,8,power_law_1.01,0.06159103870391845
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,4,8,balanced,0.08060671806335448
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,4,8,power_law_1.01,0.1097702407836914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,4,8,power_law_1.01,0.10891648292541505
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,4,8,power_law_1.01,0.10931072235107422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,4,8,power_law_1.01,0.11024383544921874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,4,8,power_law_1.01,0.11016575813293457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,4,8,power_law_1.01,0.1130777645111084
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,4,8,power_law_1.01,0.11512191772460938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,4,8,power_law_1.01,0.11436544418334962
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,4,8,power_law_1.01,0.11789952278137208
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,4,8,power_law_1.01,0.11869695663452148
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,4,8,power_law_1.01,0.11977215766906739
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,4,8,power_law_1.01,0.12411775588989257
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,4,8,power_law_1.01,0.12716927528381347
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,4,8,power_law_1.01,0.1500326442718506
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,4,8,power_law_1.01,0.16755584716796873
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,4,8,power_law_1.01,0.1869990348815918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,4,8,power_law_1.01,0.28508159637451175
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,4,8,power_law_1.01,0.35413887023925783
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,4,8,power_law_1.01,0.45737342834472655
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,16,1,power_law_1.2,18.793782958984373
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,4,8,power_law_1.01,0.6438706970214844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,4,8,power_law_1.01,0.9291212463378906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,4,8,power_law_1.01,1.1929676818847657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,16,1,power_law_1.2,25.0927490234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,4,8,power_law_1.01,0.03680255889892578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,4,8,power_law_1.01,1.856248321533203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,4,8,power_law_1.01,0.05589888095855713
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,4,8,power_law_1.01,0.05689856052398682
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,4,8,power_law_1.01,0.056570878028869634
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,4,8,power_law_1.01,0.05731584072113037
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,4,8,power_law_1.01,0.05785215854644775
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,4,8,power_law_1.01,0.05944704055786133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,4,8,power_law_1.01,0.06008831977844238
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,4,8,power_law_1.01,0.06079232215881347
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,4,8,power_law_1.01,0.062186241149902344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,4,8,power_law_1.01,0.06420095920562743
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,4,8,power_law_1.01,0.0658022403717041
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,4,8,power_law_1.01,0.06772992134094238
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,4,8,power_law_1.01,0.07078911781311034
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,4,8,power_law_1.01,0.07268991947174072
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,4,8,power_law_1.01,0.07905151844024658
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,4,8,power_law_1.01,0.07850368022918701
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,4,8,power_law_1.01,0.112805118560791
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,4,8,power_law_1.01,0.10754431724548338
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,4,8,power_law_1.01,0.16979328155517578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,4,8,power_law_1.01,2.4977267456054686
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,4,8,power_law_1.01,4.426644592285156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,4,8,power_law_1.01,0.213809928894043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,4,8,power_law_1.01,3.8297445678710935
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,4,8,power_law_1.01,0.29698944091796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,16,1,power_law_1.2,30.62571044921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,4,8,power_law_1.01,0.3718067169189453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,4,8,power_law_1.2,0.10862976074218751
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,4,8,power_law_1.2,0.10989439964294434
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,4,8,power_law_1.2,0.10950912475585936
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,4,8,power_law_1.2,0.11026432037353515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,4,8,power_law_1.2,0.11135744094848632
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,4,8,power_law_1.2,0.11314175605773927
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,4,8,power_law_1.2,0.11525888442993164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,4,8,power_law_1.2,0.11579903602600097
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,4,8,power_law_1.2,0.11657471656799316
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,4,8,power_law_1.2,0.06243328094482422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,4,8,power_law_1.2,0.11929216384887695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,4,8,power_law_1.2,0.12163455963134766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,4,8,power_law_1.2,0.12272768020629883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,4,8,power_law_1.2,0.12813695907592773
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,4,8,power_law_1.2,0.15182720184326173
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,4,8,power_law_1.2,0.17007999420166015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,4,8,power_law_1.2,0.1967987251281738
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,4,8,power_law_1.2,0.35706497192382813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,4,8,power_law_1.2,0.2971379280090332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,4,8,power_law_1.2,0.5075199890136719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,4,8,power_law_1.01,0.5329484939575195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,4,8,power_law_1.2,0.707930908203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,4,8,power_law_1.01,0.7468851470947266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,4,8,power_law_1.01,1.204373779296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,4,8,power_law_1.01,1.5671859741210938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,4,8,power_law_1.2,0.9877542114257812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,4,8,power_law_1.2,0.03600512027740478
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,4,8,power_law_1.2,0.0560972785949707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,4,8,power_law_1.2,0.056736001968383784
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,4,8,power_law_1.2,0.056710400581359864
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,4,8,power_law_1.2,0.05676032066345214
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,4,8,power_law_1.2,0.05704063892364501
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,4,8,power_law_1.2,0.05822847843170166
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,4,8,power_law_1.2,0.06020351886749268
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,4,8,power_law_1.2,0.061272320747375486
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,4,8,power_law_1.2,1.2089702606201171
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,4,8,power_law_1.2,0.06190591812133789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,4,8,power_law_1.2,0.0645350408554077
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,4,8,power_law_1.2,0.06602880001068115
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,4,8,power_law_1.2,0.06739071846008302
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,4,8,power_law_1.2,0.07132287979125976
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,4,8,power_law_1.2,0.07423999786376953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,4,8,power_law_1.2,0.07738368034362793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,4,8,power_law_1.2,0.0787443208694458
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,4,8,power_law_1.2,1.8959014892578125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,4,8,power_law_1.2,0.11252991676330566
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,4,8,power_law_1.2,0.10715519905090332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,4,8,power_law_1.2,0.17054975509643555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,4,8,power_law_1.2,0.22331775665283202
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,4,8,power_law_1.2,0.29939712524414064
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,16,1,power_law_1.2,43.261181640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,4,8,power_law_1.2,0.36814849853515624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,4,8,power_law_1.2,2.6516824340820313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,8,4,balanced,0.04508416175842285
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,8,4,balanced,0.04357759952545166
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,8,4,balanced,0.05248256206512451
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,8,4,balanced,0.08832639694213867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,4,8,power_law_1.2,0.5600985717773438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,8,4,balanced,0.08889216423034668
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,8,4,balanced,0.08855168342590332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,8,4,balanced,0.0901427173614502
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,8,4,balanced,0.08993280410766602
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,8,4,balanced,0.09285120010375977
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,8,4,balanced,0.09248255729675293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,8,4,balanced,0.0933516788482666
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,8,4,balanced,0.09497088432312012
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,8,4,balanced,0.09656319618225098
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,8,4,balanced,0.09773440361022949
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,8,4,balanced,0.09990783691406249
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,8,4,balanced,0.10058879852294922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,8,4,balanced,0.10503680229187011
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,8,4,balanced,0.11674367904663085
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,8,4,balanced,0.11103487968444825
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,8,4,balanced,0.16577791213989257
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,8,4,balanced,0.21198335647583008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,8,4,balanced,0.30955135345458984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,8,4,balanced,0.38135040283203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,8,4,balanced,0.547212791442871
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,4,8,power_law_1.2,4.093382263183594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,8,4,balanced,0.7203327941894531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,8,4,balanced,0.027029759883880615
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,8,4,balanced,0.02692352056503296
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,8,4,balanced,0.033149440288543705
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,8,4,balanced,1.0566758728027343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,8,4,balanced,0.04851071834564209
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,8,4,balanced,0.04868607997894287
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,8,4,balanced,0.0485427188873291
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,8,4,balanced,0.04910592079162598
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,8,4,balanced,1.4013209533691406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,8,4,balanced,0.04915328025817871
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,8,4,balanced,0.04956799983978272
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,8,4,balanced,0.051866879463195806
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,8,4,balanced,0.05157248020172119
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,8,4,balanced,0.05255551815032959
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,8,4,balanced,0.05470848083496094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,8,4,balanced,0.05550335884094239
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,8,4,balanced,0.05763455867767334
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,8,4,balanced,0.06190847873687744
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,8,4,balanced,0.0677849578857422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,8,4,balanced,0.049943041801452634
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,8,4,balanced,0.06219007968902588
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,8,4,balanced,0.07092351913452148
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,8,4,balanced,0.08509951591491699
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,8,4,balanced,0.12274944305419921
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,8,4,balanced,0.14544896125793455
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,8,4,balanced,0.20839551925659178
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,8,4,balanced,0.27088895797729495
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,8,4,balanced,0.3937100982666016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,8,4,power_law_1.01,0.04549888134002685
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,8,4,balanced,0.5038463973999023
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,8,4,power_law_1.01,0.04914559841156006
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,8,4,power_law_1.01,0.08651647567749024
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,8,4,power_law_1.01,0.08798080444335937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,4,8,power_law_1.2,0.7868800354003906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,8,4,power_law_1.01,0.08958080291748047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,8,4,power_law_1.01,0.08966272354125977
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,8,4,power_law_1.01,0.09105152130126953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,8,4,power_law_1.01,0.09413375854492187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,8,4,power_law_1.01,0.09485952377319336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,8,4,power_law_1.01,0.09567359924316407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,8,4,power_law_1.01,0.09833087921142578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,8,4,power_law_1.01,0.09820159912109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,8,4,power_law_1.01,0.10004096031188965
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,8,4,power_law_1.01,0.10360192298889159
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,8,4,power_law_1.01,0.11189248085021972
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,8,4,power_law_1.01,0.11813632011413575
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,8,4,power_law_1.01,0.13409536361694335
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,8,4,power_law_1.01,0.22144384384155275
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,8,4,power_law_1.01,0.19548927307128908
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,8,4,power_law_1.01,0.2861939239501953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,8,4,power_law_1.01,0.47088897705078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,8,4,power_law_1.01,0.5826265716552734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,8,4,power_law_1.01,0.8380108642578126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,8,4,power_law_1.01,1.0462438201904296
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,8,4,power_law_1.01,0.03078783988952637
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,8,4,power_law_1.01,0.031896319389343265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,8,4,power_law_1.01,0.0466918420791626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,8,4,power_law_1.01,0.04769279956817627
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,8,4,power_law_1.01,0.04872960090637207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,8,4,power_law_1.01,0.04862080097198486
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,8,4,power_law_1.01,0.05026048183441162
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,8,4,power_law_1.01,0.05042943954467773
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,8,4,power_law_1.01,0.05212031841278077
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,8,4,power_law_1.01,0.05212031841278077
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,8,4,power_law_1.01,0.05450751781463623
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,8,4,power_law_1.01,0.056156158447265625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,4,8,power_law_1.2,5.465767822265625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,8,4,power_law_1.01,0.05807744026184082
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,8,4,power_law_1.01,1.6260134887695312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,8,4,power_law_1.01,0.0625267219543457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,8,4,power_law_1.01,0.06463871955871582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,4,8,power_law_1.2,1.2050713348388673
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,8,4,power_law_1.01,0.06752639770507812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,8,4,power_law_1.01,0.06483583927154542
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,8,4,power_law_1.01,0.09798399925231933
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,8,4,power_law_1.01,0.1274726390838623
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,8,4,power_law_1.01,0.08837120056152344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,16,1,power_law_1.2,54.754482421875004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,8,4,power_law_1.01,0.48979198455810546
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,8,4,power_law_1.01,0.1662668800354004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,8,4,power_law_1.01,0.22214271545410158
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,8,4,power_law_1.2,0.04548480033874512
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,8,4,power_law_1.2,0.04903679847717285
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,8,4,power_law_1.2,0.08594176292419434
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,8,4,power_law_1.01,0.27133056640625003
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,8,4,power_law_1.2,0.08828543663024903
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,8,4,power_law_1.2,0.08855936050415039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,8,4,power_law_1.2,0.08989055633544922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,8,4,power_law_1.2,0.09194111824035645
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,8,4,power_law_1.2,0.09331456184387207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,8,4,power_law_1.2,0.09481599807739258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,8,4,power_law_1.2,0.09596159934997558
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,8,4,power_law_1.2,0.0980339241027832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,8,4,power_law_1.2,0.09920639991760254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,8,4,power_law_1.2,0.10018048286437989
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,8,4,power_law_1.2,0.10573696136474608
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,8,4,power_law_1.2,0.11070848464965821
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,8,4,power_law_1.2,0.12115584373474123
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,8,4,power_law_1.2,0.1389516830444336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,8,4,power_law_1.2,0.23260671615600587
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,8,4,power_law_1.2,0.20559104919433593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,8,4,power_law_1.01,0.4000947189331055
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,8,4,power_law_1.01,2.650543212890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,4,8,power_law_1.2,1.6731494140624998
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,8,4,power_law_1.2,0.34463230133056644
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,8,4,power_law_1.2,0.609452781677246
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,8,4,power_law_1.2,0.4732799911499024
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,8,4,power_law_1.2,0.7519513702392577
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,8,4,power_law_1.2,0.029885439872741698
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,8,4,power_law_1.2,0.031192319393157957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,8,4,power_law_1.2,0.04688767910003662
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,8,4,power_law_1.01,0.8607347106933594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,8,4,power_law_1.2,0.0477785587310791
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,8,4,power_law_1.2,0.052801280021667486
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,8,4,power_law_1.2,0.05273344039916992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,8,4,power_law_1.01,2.1409983825683594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,8,4,power_law_1.2,0.049446401596069334
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,8,4,power_law_1.2,0.05036032199859619
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,8,4,power_law_1.2,0.050446081161499026
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,8,4,power_law_1.2,0.05219967842102051
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,8,4,power_law_1.2,0.05330687999725342
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,8,4,power_law_1.2,0.056468482017517085
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,8,4,power_law_1.2,0.058106880187988284
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,8,4,power_law_1.2,0.0617907190322876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,8,4,power_law_1.2,0.06403584003448486
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,8,4,power_law_1.2,0.06713088035583496
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,8,4,power_law_1.2,0.06511231899261474
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,8,4,power_law_1.2,1.4825112915039063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,8,4,power_law_1.2,0.09841279983520508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,8,4,power_law_1.2,0.0840998363494873
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,8,4,power_law_1.2,0.13030400276184081
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,8,4,power_law_1.2,0.2208639907836914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,8,4,power_law_1.2,0.17011072158813476
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,8,4,power_law_1.2,1.1548365020751954
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,8,4,power_law_1.01,1.0053056335449218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,8,4,power_law_1.2,0.2625996780395508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,16,2,balanced,0.024958720207214354
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,16,2,balanced,0.040331521034240723
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,16,2,balanced,0.048995838165283204
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,16,2,balanced,0.07852159976959229
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,16,2,balanced,0.07935616016387939
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,16,2,balanced,0.07996032238006592
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,16,2,balanced,0.08222975730895996
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,16,2,balanced,0.08294655799865722
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,16,2,balanced,0.08401023864746093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,16,2,balanced,0.08484352111816407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,16,2,balanced,0.0867251205444336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,16,2,balanced,0.08770815849304199
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,16,2,balanced,0.08958975791931152
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,16,2,balanced,0.09338624000549316
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,16,2,balanced,0.09881343841552734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,16,2,balanced,0.1010086441040039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,16,2,balanced,0.09574912071228028
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,16,2,balanced,0.11043583869934082
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,16,2,balanced,0.12464768409729003
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,16,2,balanced,0.1701568031311035
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,16,2,balanced,0.22333696365356442
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,16,2,balanced,0.33413246154785153
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,16,2,balanced,0.40957054138183596
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,16,2,balanced,0.5918988800048828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,16,2,balanced,0.7780863952636718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,16,2,balanced,1.136563186645508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,16,2,balanced,0.024993278980255128
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,16,2,balanced,0.02648576021194458
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,16,2,balanced,0.03190144062042236
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,16,2,balanced,1.5013337707519532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,16,2,balanced,0.044605441093444824
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,16,2,balanced,0.04429696083068847
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,8,4,power_law_1.2,0.4035404968261719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,16,2,balanced,0.04536064147949219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,16,2,balanced,0.046603522300720214
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,16,2,balanced,0.04707968235015869
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,16,2,balanced,0.04775551795959473
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,16,2,balanced,0.050000638961791986
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,16,2,balanced,0.05126016139984131
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,16,2,balanced,0.05328383922576905
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,16,2,balanced,0.056943359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,16,2,balanced,0.04326272010803223
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,16,2,balanced,0.06041344165802002
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,16,2,balanced,0.06301695823669433
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,16,2,balanced,0.05689343929290771
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,16,2,balanced,0.06708223819732666
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,16,2,balanced,0.07347968101501465
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,16,2,balanced,0.08805248260498047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,16,2,balanced,0.10628607749938965
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,16,2,balanced,0.15844736099243165
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,16,2,balanced,0.18946943283081055
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,16,2,balanced,0.27600128173828126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,16,2,balanced,0.36262657165527346
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,16,2,power_law_1.01,0.038684160709381105
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,16,2,balanced,0.537844467163086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,16,2,power_law_1.01,0.04143487930297852
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,16,2,power_law_1.01,0.043968000411987306
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,16,2,balanced,0.6857676696777344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,16,2,power_law_1.01,0.07908864021301269
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,8,4,power_law_1.2,0.5099699020385742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,16,2,power_law_1.01,0.07989247798919677
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,16,2,power_law_1.01,0.08035584449768066
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,16,2,power_law_1.01,0.08226176261901855
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,16,2,power_law_1.01,0.0828825569152832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,16,2,power_law_1.01,0.08436736106872558
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,16,2,power_law_1.01,0.08491135597229003
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,16,2,power_law_1.01,0.08697088241577149
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,16,2,power_law_1.01,0.0882316780090332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,16,2,power_law_1.01,0.09049856185913085
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,16,2,power_law_1.01,0.09661696434020997
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,16,2,power_law_1.01,0.10435199737548828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,16,2,power_law_1.01,0.10902400016784668
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,16,2,power_law_1.01,0.14817791938781738
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,16,2,power_law_1.01,0.13133312225341798
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,16,2,power_law_1.01,0.1050534439086914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,16,2,power_law_1.01,0.21242752075195312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,16,2,power_law_1.01,0.30774272918701173
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,16,2,power_law_1.01,0.553834228515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,16,2,power_law_1.01,0.39165439605712893
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,16,2,power_law_1.01,0.7587596893310546
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,16,2,power_law_1.01,0.02547327995300293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,16,2,power_law_1.01,0.02888832092285156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,16,2,power_law_1.01,0.0301529598236084
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,16,2,power_law_1.01,0.04360320091247559
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,16,2,power_law_1.01,0.044279041290283205
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,8,4,power_law_1.2,3.264551696777344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,16,2,power_law_1.01,0.045633277893066405
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,16,2,power_law_1.01,0.04717567920684814
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,16,2,power_law_1.01,0.04620800018310547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,16,2,power_law_1.01,0.04842495918273926
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,16,2,power_law_1.01,0.04806528091430664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,16,2,power_law_1.01,0.05120512008666992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,16,2,power_law_1.01,0.05237504005432129
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,16,2,power_law_1.01,0.05382400035858155
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,16,2,power_law_1.01,0.05805568218231201
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,16,2,power_law_1.01,0.06141568183898925
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,16,2,power_law_1.01,0.06385536193847656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,8,4,power_law_1.2,0.8372966766357421
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,16,2,power_law_1.01,0.06137983798980713
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,16,2,power_law_1.01,0.07646463871002197
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,16,2,power_law_1.01,0.06915328025817871
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,16,2,power_law_1.01,0.11225600242614746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,16,2,power_law_1.01,0.13602304458618164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,16,2,power_law_1.01,0.1920345687866211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,16,2,power_law_1.01,1.0792384338378906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,8,4,power_law_1.2,1.0553919982910156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,8,4,power_law_1.2,2.164752655029297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,16,2,power_law_1.2,0.03936000108718872
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,16,2,power_law_1.2,0.041011199951171876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,16,2,power_law_1.2,0.043576321601867675
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,16,2,power_law_1.2,0.078886399269104
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,16,2,power_law_1.2,0.07919104099273681
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,16,2,power_law_1.2,0.0810483169555664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,16,2,power_law_1.2,0.08214783668518066
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,16,2,power_law_1.2,0.08310655593872071
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,16,2,power_law_1.2,0.08366720199584961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,16,2,power_law_1.2,0.08468607902526856
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,16,2,power_law_1.2,0.08670207977294922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,16,2,power_law_1.2,0.08810879707336425
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,16,2,power_law_1.2,0.09042176246643066
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,16,2,power_law_1.2,0.09837823867797851
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,16,2,power_law_1.2,0.10231807708740234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,16,2,power_law_1.01,0.2368179130554199
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,16,2,power_law_1.2,0.10652031898498535
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,16,2,power_law_1.01,0.32313343048095705
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,16,2,power_law_1.2,0.13347328186035157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,16,2,power_law_1.2,0.16450048446655274
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,16,2,power_law_1.2,0.11248767852783204
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,16,2,power_law_1.2,0.23826431274414062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,16,2,power_law_1.01,0.41638912200927736
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,16,2,power_law_1.2,0.3155148887634277
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,16,2,power_law_1.01,1.6659507751464844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,16,2,power_law_1.2,0.5251494216918945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,16,2,power_law_1.01,1.4968103027343749
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,16,2,power_law_1.2,0.026835200786590574
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,16,2,power_law_1.2,0.4236979293823242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,16,2,power_law_1.01,0.7640819549560547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,16,2,power_law_1.2,0.028395519256591794
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,16,2,power_law_1.2,0.02886528015136719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,16,2,power_law_1.2,0.04321407794952393
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,16,2,power_law_1.2,0.04298880100250244
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,16,2,power_law_1.2,0.043715839385986326
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,16,2,power_law_1.2,0.04611455917358399
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,16,2,power_law_1.2,0.04699007987976074
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,16,2,power_law_1.2,0.04694399833679199
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,16,2,power_law_1.2,0.048890881538391114
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,16,2,power_law_1.2,0.04947455883026123
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,16,2,power_law_1.2,0.052650241851806646
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,16,2,power_law_1.2,0.0537830400466919
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,16,2,power_law_1.2,0.05798655986785889
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,16,2,power_law_1.2,0.06129536151885986
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,16,2,power_law_1.2,0.06384384155273437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,16,2,power_law_1.2,0.06351871967315674
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,16,2,power_law_1.2,0.07468031883239747
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,16,2,power_law_1.2,0.07002240180969238
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,16,2,power_law_1.2,0.11079296112060547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,16,2,power_law_1.2,1.0983513641357423
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,16,2,power_law_1.2,0.13524864196777345
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,16,2,power_law_1.01,0.613331184387207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,16,2,power_law_1.2,0.1833919906616211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,16,2,power_law_1.2,0.3162060737609863
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,32,1,balanced,0.02793152093887329
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,32,1,balanced,0.043249921798706056
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,32,1,balanced,0.05313407897949218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,32,1,balanced,0.05350592136383057
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,32,1,balanced,0.07974783897399902
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,32,1,balanced,0.08069439888000488
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,16,2,power_law_1.2,0.7386329650878907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,32,1,balanced,0.08244095802307129
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,32,1,balanced,0.08385151863098145
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,32,1,balanced,0.08402239799499511
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,32,1,balanced,0.0853657627105713
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,32,1,balanced,0.08703871726989745
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,32,1,balanced,0.08887488365173339
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,32,1,balanced,0.09003392219543457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,32,1,balanced,0.0938758373260498
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,32,1,balanced,0.09928959846496582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,16,2,power_law_1.2,0.23991680145263672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,32,1,balanced,0.10226816177368163
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,32,1,balanced,0.10049087524414062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,32,1,balanced,0.11528384208679199
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,32,1,balanced,0.1257260799407959
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,32,1,balanced,0.15282496452331543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,32,1,balanced,0.20374784469604493
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,32,1,balanced,0.31803968429565427
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,32,1,balanced,0.3984934234619141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,32,1,balanced,0.577767677307129
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,32,1,balanced,0.7573849487304687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,32,1,balanced,1.1225682830810546
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,32,1,balanced,1.4844000244140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,32,1,balanced,0.028802559375762937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,32,1,balanced,0.030692479610443114
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,32,1,balanced,0.03338624000549316
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,16,2,power_law_1.2,0.42165889739990237
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,16,2,power_law_1.2,1.9373580932617187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,32,1,balanced,0.033724160194396974
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,16,2,power_law_1.2,0.6444172668457031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,32,1,balanced,0.049142398834228516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,32,1,balanced,0.05426496028900146
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,32,1,balanced,0.05540800094604492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,32,1,balanced,0.056928639411926274
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,32,1,balanced,0.05940671920776367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,32,1,balanced,0.06202432155609131
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,32,1,balanced,0.06419775962829591
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,32,1,balanced,0.050299520492553706
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,32,1,balanced,0.0618291187286377
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,32,1,balanced,0.07174015998840332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,32,1,balanced,0.07680831909179688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,32,1,balanced,0.0925926399230957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,32,1,balanced,0.1107372760772705
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,32,1,balanced,0.16863168716430663
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,16,2,power_law_1.2,1.4724954223632811
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,32,1,balanced,0.2057088088989258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,32,1,balanced,0.2955878448486328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,32,1,balanced,0.05195583820343017
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,32,1,power_law_1.01,0.04070271968841553
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,32,1,balanced,0.05099071979522705
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,32,1,power_law_1.01,0.04503680229187011
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,32,1,power_law_1.01,0.04827904224395752
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,32,1,power_law_1.01,0.053232002258300784
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,32,1,balanced,0.7427782440185547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,32,1,power_law_1.01,0.0791916799545288
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,32,1,power_law_1.01,0.0803654384613037
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,32,1,power_law_1.01,0.08191103935241699
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,32,1,power_law_1.01,0.0829196834564209
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,32,1,balanced,0.05279744148254395
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,32,1,power_law_1.01,0.08522751808166504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,32,1,power_law_1.01,0.08822719573974609
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,32,1,power_law_1.01,0.08947327613830566
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,32,1,balanced,0.5634771347045898
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,32,1,power_law_1.01,0.09255359649658204
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,32,1,power_law_1.01,0.09543999671936035
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,32,1,power_law_1.01,0.0982630443572998
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,32,1,power_law_1.01,0.10385855674743652
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,32,1,power_law_1.01,0.1059603214263916
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,32,1,power_law_1.01,0.0838809585571289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,32,1,power_law_1.01,0.11969216346740723
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,32,1,power_law_1.01,0.13345664024353027
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,32,1,balanced,0.052965121269226076
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,32,1,power_law_1.01,0.17850048065185548
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,32,1,power_law_1.01,0.2316044807434082
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,32,1,power_law_1.01,0.3354963302612305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,32,1,balanced,0.3839411163330078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,32,1,power_law_1.01,0.4186592102050781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,32,1,power_law_1.01,0.02843712091445923
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,32,1,power_law_1.01,0.5924441528320312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,32,1,power_law_1.01,0.04962048053741455
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,32,1,power_law_1.01,0.05022592067718505
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,32,1,power_law_1.01,0.03247296094894409
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,32,1,power_law_1.01,0.0515993595123291
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,32,1,power_law_1.01,0.05206016063690185
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,32,1,power_law_1.01,0.0528275203704834
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,32,1,power_law_1.01,0.05376704216003418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,32,1,power_law_1.01,0.03554431915283203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,32,1,power_law_1.01,0.05483200073242187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,32,1,power_law_1.01,0.05619840145111084
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,32,1,power_law_1.01,0.06002175807952881
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,32,1,power_law_1.01,0.06491968154907227
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,32,1,power_law_1.01,0.06716288089752197
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,32,1,power_law_1.01,0.06915328025817871
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,32,1,power_law_1.01,0.06679679870605468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,32,1,power_law_1.01,0.7772057342529297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,32,1,power_law_1.01,0.07679232120513915
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,32,1,power_law_1.01,0.08280832290649413
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,32,1,power_law_1.01,0.09958463668823242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,32,1,power_law_1.01,0.033787519931793214
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,32,1,power_law_1.01,0.13940928459167481
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,32,1,power_law_1.01,0.17642431259155272
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,16,2,power_law_1.2,0.7677056121826171
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,32,1,power_law_1.2,0.040206079483032224
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,32,1,power_law_1.2,0.04515711784362793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,32,1,power_law_1.2,0.04646143913269043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,32,1,power_law_1.2,0.053211522102355954
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,32,1,power_law_1.01,0.2342131233215332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,32,1,power_law_1.2,0.07890240192413331
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,32,1,power_law_1.2,0.08070464134216308
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,32,1,power_law_1.2,0.08245823860168458
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,32,1,power_law_1.2,0.08319999694824218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,32,1,power_law_1.2,0.08415936470031739
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,32,1,power_law_1.2,0.08531519889831543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,32,1,power_law_1.2,0.08817279815673829
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,32,1,power_law_1.2,0.08974016189575196
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,32,1,power_law_1.01,1.1287142181396486
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,32,1,power_law_1.2,0.09136704444885255
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,32,1,power_law_1.2,0.09574272155761718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,32,1,power_law_1.01,0.3244300842285156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,32,1,power_law_1.2,0.09928000450134278
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,32,1,power_law_1.2,0.10552384376525878
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,32,1,power_law_1.2,0.10568191528320312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,32,1,power_law_1.2,0.12014911651611329
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,32,1,power_law_1.2,0.13525952339172362
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,32,1,power_law_1.01,0.39766014099121094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,32,1,power_law_1.2,0.1830246353149414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,32,1,power_law_1.2,0.2326425552368164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,32,1,power_law_1.01,1.4958828735351564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,32,1,power_law_1.2,0.334218864440918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,32,1,power_law_1.2,0.0289516806602478
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,32,1,power_law_1.2,0.03278592109680176
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,32,1,power_law_1.2,0.032927999496459956
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,32,1,power_law_1.01,0.5713459014892578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,32,1,power_law_1.2,0.048808321952819825
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,32,1,power_law_1.2,0.049498882293701175
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,32,1,power_law_1.2,0.05089791774749756
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,32,1,power_law_1.2,0.051623039245605476
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,32,1,power_law_1.2,0.052657279968261715
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,32,1,power_law_1.2,0.4094796752929687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,32,1,power_law_1.2,0.05279295921325684
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,32,1,power_law_1.2,0.05502912044525147
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,32,1,power_law_1.2,0.056466560363769534
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,32,1,power_law_1.2,0.06026368141174317
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,32,1,power_law_1.2,0.06506944179534913
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,32,1,power_law_1.2,0.06715583801269531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,32,1,power_law_1.2,0.06939648151397705
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,32,1,power_law_1.2,0.06665408134460449
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,32,1,power_law_1.2,0.03507456064224243
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,32,1,power_law_1.2,0.07636223793029785
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,32,1,power_law_1.2,0.0822815990447998
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,32,1,power_law_1.2,0.09911168098449707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,32,1,power_law_1.01,0.7542758178710938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,32,1,power_law_1.2,0.1394636821746826
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,32,1,power_law_1.2,0.5886009597778321
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,32,1,power_law_1.2,0.17600831985473633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,32,1,power_law_1.2,0.7754668426513672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,32,1,power_law_1.2,0.23447231292724607
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,4,8,balanced,0.08201600074768066
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,4,8,balanced,0.08134400367736816
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,4,8,balanced,0.08233728408813476
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,4,8,balanced,0.14744704246520995
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,4,8,balanced,0.14853887557983397
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,4,8,balanced,0.14700032234191895
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,4,8,balanced,0.14956543922424317
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,4,8,balanced,0.14877568244934083
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,4,8,balanced,0.14917759895324706
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,4,8,balanced,0.15027199745178224
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,4,8,balanced,0.15203712463378907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,4,8,balanced,0.15622528076171877
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,4,8,balanced,0.1577228832244873
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,4,8,balanced,0.1595417594909668
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,4,8,balanced,0.15844736099243165
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,4,8,balanced,0.15888511657714843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,4,8,balanced,0.16953855514526367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,4,8,balanced,0.17220735549926758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,4,8,balanced,0.17589759826660156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,4,8,balanced,0.30810367584228515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,4,8,balanced,0.3343334579467773
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,4,8,balanced,0.47959041595458984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,4,8,balanced,0.6324787139892578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,4,8,balanced,0.896579818725586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,4,8,balanced,1.1798336029052734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,32,1,power_law_1.2,0.3251059341430664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,4,8,balanced,0.04883967876434327
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,4,8,balanced,0.04859903812408447
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,4,8,balanced,1.7673536682128905
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,4,8,balanced,0.048876800537109376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,4,8,balanced,0.07428864002227784
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,4,8,balanced,0.07002495765686036
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,4,8,balanced,0.07309311866760254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,4,8,balanced,0.07061376094818114
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,4,8,balanced,0.07300352096557618
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,4,8,balanced,0.07120384216308594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,4,8,balanced,0.07343232154846191
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,4,8,balanced,0.07155072212219238
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,4,8,balanced,0.0730406379699707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,4,8,balanced,0.07257343769073486
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,4,8,balanced,2.3190681457519533
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,4,8,balanced,0.07410048007965088
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,4,8,balanced,0.08027008056640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,4,8,balanced,0.0783244800567627
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,4,8,balanced,0.08531455993652344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,4,8,balanced,0.08776576042175292
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,4,8,balanced,0.10096896171569825
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,4,8,balanced,0.10222720146179198
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,4,8,balanced,0.16388736724853517
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,4,8,balanced,0.18273151397705079
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,4,8,balanced,0.25863040924072267
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,4,8,balanced,0.07453055858612061
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,4,8,balanced,0.32131584167480465
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,4,8,balanced,0.48557056427001954
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,4,8,balanced,0.6248908615112304
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,4,8,power_law_1.01,0.0816806411743164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,4,8,power_law_1.01,0.14544511795043946
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,4,8,power_law_1.01,0.1477235221862793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,4,8,power_law_1.01,0.1474764823913574
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,4,8,power_law_1.01,0.14864895820617677
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,4,8,power_law_1.01,0.1503219223022461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,4,8,power_law_1.01,0.1545792007446289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,4,8,power_law_1.01,0.1577254390716553
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,4,8,power_law_1.01,0.15627391815185548
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,4,8,power_law_1.01,0.15715968132019043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,4,8,power_law_1.01,0.16385919570922852
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,4,8,power_law_1.01,0.16450559616088867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,4,8,power_law_1.01,0.16698368072509767
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,32,1,power_law_1.2,0.40001663208007815
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,4,8,power_law_1.01,0.1725107192993164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,4,8,power_law_1.01,0.20523008346557617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,4,8,power_law_1.01,0.2669452857971192
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,4,8,power_law_1.01,0.3137612724304199
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,4,8,power_law_1.01,0.4241484832763672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,4,8,power_law_1.01,0.527850227355957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,32,1,power_law_1.2,1.132135009765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,4,8,power_law_1.01,0.8601599884033202
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,4,8,power_law_1.01,1.1665369415283204
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,4,8,power_law_1.01,1.522097930908203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,4,8,power_law_1.01,3.1309747314453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,4,8,power_law_1.01,0.05029632091522217
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,4,8,power_law_1.01,0.0714035177230835
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,4,8,power_law_1.01,2.2336627197265626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,4,8,power_law_1.01,0.07164927959442138
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,4,8,power_law_1.01,0.07271423816680908
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,4,8,power_law_1.01,0.0717516803741455
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,4,8,power_law_1.01,0.07264256000518798
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,4,8,power_law_1.01,0.07309696197509766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,4,8,power_law_1.01,0.07378560066223144
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,4,8,power_law_1.01,0.07469567775726318
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,4,8,power_law_1.01,0.0753382396697998
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,4,8,power_law_1.01,0.07682559967041017
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,4,8,power_law_1.01,0.07963776111602783
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,4,8,power_law_1.01,0.08165375709533691
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,4,8,power_law_1.01,0.08497407913208008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,4,8,power_law_1.01,0.09128576278686523
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,4,8,power_law_1.01,0.09921407699584961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,4,8,power_law_1.01,0.10396544456481933
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,4,8,power_law_1.01,0.16214399337768554
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,4,8,power_law_1.01,5.172324829101562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,4,8,power_law_1.01,0.17094528198242187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,4,8,power_law_1.01,0.30455551147460935
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,4,8,power_law_1.01,0.2359872055053711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,32,1,power_law_1.2,0.5738124847412109
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,4,8,power_law_1.01,0.4600998306274414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,4,8,power_law_1.01,8.8531494140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,32,1,power_law_1.2,1.4987788391113281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,4,8,power_law_1.2,0.08228863716125488
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,4,8,power_law_1.2,0.14676992416381834
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,4,8,power_law_1.2,0.1483187198638916
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,4,8,power_law_1.2,0.14846719741821288
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,4,8,power_law_1.01,1.1915392303466796
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,4,8,power_law_1.2,0.14846464157104494
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,4,8,power_law_1.2,0.1578649616241455
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,4,8,power_law_1.2,0.1544166374206543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,4,8,power_law_1.01,0.5723136138916016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,4,8,power_law_1.2,0.16057727813720704
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,4,8,power_law_1.2,0.15635840415954588
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,4,8,power_law_1.2,0.15669504165649414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,4,8,power_law_1.2,0.1670412826538086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,4,8,power_law_1.01,0.885076446533203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,4,8,power_law_1.2,0.1645721626281738
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,4,8,power_law_1.01,4.1615576171874995
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,4,8,power_law_1.2,0.1701260757446289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,4,8,power_law_1.2,0.17461759567260743
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,4,8,power_law_1.2,0.23979007720947268
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,4,8,power_law_1.2,0.3399347305297852
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,4,8,power_law_1.2,0.29427328109741213
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,4,8,power_law_1.2,0.43537025451660155
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,4,8,power_law_1.01,2.4999436950683593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,4,8,power_law_1.2,0.5421452713012695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,4,8,power_law_1.2,0.8512947082519531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,4,8,power_law_1.2,1.0714367675781251
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,4,8,power_law_1.2,2.161085510253906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,4,8,power_law_1.2,0.04930047988891602
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,4,8,power_law_1.2,0.07148159980773926
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,4,8,power_law_1.2,0.07132800102233887
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,4,8,power_law_1.2,0.07093376159667969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,4,8,power_law_1.2,0.07120895862579346
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,4,8,power_law_1.2,0.07184512138366698
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,4,8,power_law_1.2,0.07260672092437745
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,4,8,power_law_1.2,0.07347839832305908
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,4,8,power_law_1.2,0.07451136112213134
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,4,8,power_law_1.2,0.0751961612701416
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,4,8,power_law_1.2,0.07714560031890869
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,4,8,power_law_1.2,0.07982848167419435
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,4,8,power_law_1.2,0.08261887550354004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,4,8,power_law_1.2,1.7191258239746094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,4,8,power_law_1.2,0.08659711837768555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,4,8,power_law_1.2,0.09218175888061524
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,32,1,power_law_1.2,0.7549651336669922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,4,8,power_law_1.2,0.10445952415466309
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,4,8,power_law_1.2,0.09752832412719727
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,4,8,power_law_1.2,0.15179519653320311
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,4,8,power_law_1.01,1.6502182006835937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,4,8,power_law_1.2,0.17290624618530273
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,4,8,power_law_1.2,0.2417318344116211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,4,8,power_law_1.2,0.32168193817138674
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,4,8,power_law_1.2,6.858038940429688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,4,8,power_law_1.2,3.433232727050781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,4,8,power_law_1.2,4.33478271484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,8,4,balanced,0.06270080089569093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,8,4,balanced,0.06221824169158936
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,8,4,balanced,0.07218175888061523
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,8,4,balanced,0.1231116771697998
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,4,8,power_law_1.2,0.47430145263671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,8,4,balanced,0.12299136161804199
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,8,4,balanced,0.12418175697326661
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,8,4,balanced,0.1249407958984375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,8,4,balanced,0.12537856101989747
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,8,4,balanced,0.12815872192382813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,8,4,balanced,0.1284812831878662
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,8,4,balanced,0.13000191688537596
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,8,4,balanced,0.1324454402923584
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,8,4,balanced,0.1344652843475342
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,8,4,balanced,0.1349503993988037
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,8,4,balanced,0.13598976135253907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,8,4,balanced,0.13774847984313965
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,8,4,balanced,0.1448844814300537
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,8,4,balanced,0.16939647674560548
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,8,4,balanced,0.17294464111328126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,8,4,balanced,0.27273599624633793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,4,8,power_law_1.2,8.874722290039063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,8,4,balanced,0.335681266784668
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,8,4,balanced,0.4948659133911133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,8,4,balanced,0.6362521743774414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,8,4,balanced,0.9213887786865234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,4,8,power_law_1.2,0.5707929611206055
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,8,4,balanced,0.03410943984985352
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,8,4,balanced,0.032971520423889164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,8,4,balanced,1.2142733001708985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,8,4,balanced,0.06336256027221679
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,8,4,balanced,0.06228352069854737
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,8,4,balanced,0.06221951961517334
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,8,4,balanced,0.06234752178192139
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,8,4,balanced,1.8110426330566405
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,8,4,balanced,0.06196991920471191
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,8,4,balanced,0.07382912158966064
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,8,4,balanced,0.06254079818725586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,8,4,balanced,0.04600831985473633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,8,4,balanced,0.06975872039794921
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,8,4,balanced,0.06405248165130614
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,8,4,balanced,0.06515967845916748
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,8,4,balanced,0.07331456184387207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,8,4,balanced,0.0692416000366211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,8,4,balanced,0.07116672039031982
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,8,4,balanced,0.07859327793121337
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,8,4,balanced,2.403626251220703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,8,4,balanced,0.0840447998046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,8,4,balanced,0.07916287899017334
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,8,4,balanced,0.10321279525756837
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,8,4,balanced,0.12197759628295898
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,8,4,balanced,0.1750284767150879
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,8,4,balanced,0.2075814437866211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,8,4,balanced,0.3065139198303223
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,8,4,balanced,0.38822017669677733
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,8,4,power_law_1.01,0.06347392082214356
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,8,4,power_law_1.01,0.06806911945343017
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,8,4,power_law_1.01,0.1183296012878418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,8,4,power_law_1.01,0.12253567695617676
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,8,4,balanced,0.7581439971923828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,8,4,power_law_1.01,0.12370431900024415
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,8,4,power_law_1.01,0.12504063606262208
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,8,4,power_law_1.01,0.12965888023376465
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,8,4,power_law_1.01,0.13279871940612792
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,8,4,power_law_1.01,0.13564031600952148
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,8,4,power_law_1.01,0.1366592025756836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,8,4,power_law_1.01,0.1379903984069824
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,8,4,balanced,0.5702336120605469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,8,4,power_law_1.01,0.13768832206726073
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,8,4,power_law_1.01,0.1465228843688965
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,8,4,power_law_1.01,0.15887231826782228
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,8,4,power_law_1.01,0.12999168395996094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,8,4,power_law_1.01,0.1809062385559082
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,8,4,power_law_1.01,0.22096384048461912
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,4,8,power_law_1.2,0.9362342071533203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,4,8,power_law_1.2,1.8790873718261718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,8,4,power_law_1.01,0.2871232032775879
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,8,4,power_law_1.01,0.41225471496582033
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,4,8,power_law_1.2,2.425511627197266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,8,4,power_law_1.01,0.5692748641967773
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,8,4,power_law_1.01,0.834060821533203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,8,4,power_law_1.01,0.039470078945159914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,8,4,power_law_1.01,0.043114237785339356
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,8,4,power_law_1.01,0.06132351875305175
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,8,4,power_law_1.01,0.06139776229858398
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,8,4,power_law_1.01,0.06113152027130127
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,8,4,power_law_1.01,1.9104396057128905
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,8,4,power_law_1.01,0.06209536075592041
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,8,4,power_law_1.01,0.06600192070007324
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,8,4,power_law_1.01,0.06342144012451172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,8,4,power_law_1.01,0.06587135791778564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,8,4,power_law_1.01,0.06531583786010742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,8,4,power_law_1.01,0.0678656005859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,4,8,power_law_1.2,1.2141465759277343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,8,4,power_law_1.01,0.06924672126770019
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,8,4,power_law_1.01,0.08029312133789063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,8,4,power_law_1.01,0.07524352073669434
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,8,4,power_law_1.01,0.9519475555419922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,8,4,power_law_1.01,0.08105088233947753
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,8,4,power_law_1.01,1.255773468017578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,8,4,power_law_1.01,0.08394495964050293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,8,4,power_law_1.01,0.08518527984619141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,8,4,power_law_1.01,0.14681983947753907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,8,4,power_law_1.01,0.10534527778625488
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,8,4,power_law_1.01,0.17249664306640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,8,4,power_law_1.01,0.23875072479248044
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,8,4,power_law_1.01,2.5639053344726563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,8,4,power_law_1.01,0.31231872558593754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,8,4,power_law_1.2,0.06296192169189453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,8,4,power_law_1.01,0.39179649353027346
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,8,4,power_law_1.2,0.06611328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,8,4,power_law_1.2,0.12023167610168457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,8,4,power_law_1.2,0.12268159866333009
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,8,4,power_law_1.2,0.1234342384338379
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,8,4,power_law_1.2,0.12556544303894043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,8,4,power_law_1.2,0.12792832374572755
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,8,4,power_law_1.2,0.13013888359069825
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,8,4,power_law_1.2,0.13305983543395997
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,8,4,power_law_1.2,0.13435775756835938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,8,4,power_law_1.2,0.13638015747070314
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,8,4,power_law_1.2,0.13690752029418946
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,8,4,power_law_1.2,0.13805312156677246
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,8,4,power_law_1.2,0.14504192352294923
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,8,4,power_law_1.2,0.16226560592651368
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,8,4,power_law_1.2,0.18267520904541015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,8,4,power_law_1.2,0.2329523277282715
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,8,4,power_law_1.2,0.28515968322753904
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,8,4,power_law_1.2,0.41355648040771487
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,8,4,power_law_1.01,0.6995520019531251
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,8,4,power_law_1.2,0.5691827011108399
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,8,4,power_law_1.2,0.8573426818847656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,8,4,power_law_1.01,0.5953945541381837
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,8,4,power_law_1.2,0.9656422424316407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,8,4,power_law_1.01,5.216468505859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,8,4,power_law_1.01,3.441104736328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,8,4,power_law_1.2,0.03914752006530762
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,8,4,power_law_1.2,0.04316160202026367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,8,4,power_law_1.2,0.06149631977081299
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,8,4,power_law_1.2,0.061560320854187014
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,8,4,power_law_1.2,0.06155136108398438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,8,4,power_law_1.2,0.06208640098571777
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,8,4,power_law_1.2,0.06318463802337645
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,8,4,power_law_1.2,0.06407296180725097
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,8,4,power_law_1.2,0.06448383808135986
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,8,4,power_law_1.01,1.6334527587890626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,8,4,power_law_1.2,0.06550655841827394
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,8,4,power_law_1.2,1.823424072265625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,8,4,power_law_1.2,0.07326208114624023
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,8,4,power_law_1.2,0.06960256099700927
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,8,4,power_law_1.2,0.07559167861938476
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,8,4,power_law_1.2,1.3257574462890624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,8,4,power_law_1.2,0.07556863784790038
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,8,4,power_law_1.2,0.08040191650390624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,8,4,power_law_1.2,0.0832652759552002
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,8,4,power_law_1.01,1.0920793914794922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,8,4,power_law_1.2,0.11075712203979493
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,8,4,power_law_1.2,0.08964351654052735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,8,4,power_law_1.2,0.14411392211914062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,8,4,power_law_1.2,0.1710963249206543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,8,4,power_law_1.2,0.24257280349731447
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,8,4,power_law_1.2,0.3723865509033203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,8,4,power_law_1.2,0.31730815887451175
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,16,2,balanced,0.04195456027984619
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,8,4,power_law_1.2,1.9706495666503905
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,16,2,balanced,0.05741312026977539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,16,2,balanced,0.06993023872375488
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,16,2,balanced,0.11279232025146484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,16,2,balanced,0.11465855598449708
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,16,2,balanced,0.11444479942321777
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,16,2,balanced,0.11602432250976562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,16,2,balanced,0.11744000434875487
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,16,2,balanced,0.11722751617431641
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,16,2,balanced,0.11803775787353517
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,16,2,balanced,0.11880448341369629
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,16,2,balanced,0.12082176208496094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,16,2,balanced,0.12433664321899414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,16,2,balanced,0.1289139175415039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,16,2,balanced,0.13839360237121584
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,16,2,balanced,0.1426854419708252
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,16,2,balanced,0.14040960311889647
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,16,2,balanced,0.1629542350769043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,16,2,balanced,0.18228096008300781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,16,2,balanced,0.3101401519775391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,16,2,balanced,0.35237247467041016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,16,2,balanced,0.5101529693603515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,16,2,balanced,0.6631884765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,16,2,balanced,0.9623538970947265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,16,2,balanced,1.2693580627441405
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,16,2,balanced,1.897902069091797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,16,2,balanced,0.02924799919128418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,16,2,balanced,0.03173504114151001
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,16,2,balanced,0.04476160049438477
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,8,4,power_law_1.2,0.5742092895507812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,16,2,balanced,0.05740543842315674
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,16,2,balanced,2.534781494140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,16,2,balanced,0.057459201812744144
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,16,2,balanced,0.06033664226531983
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,16,2,balanced,0.05888768196105957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,16,2,balanced,0.06180223941802978
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,16,2,balanced,0.06030464172363281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,16,2,balanced,0.061761279106140134
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,16,2,balanced,0.06331647872924805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,16,2,balanced,0.06568960189819337
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,16,2,balanced,0.06699007987976074
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,16,2,balanced,0.0717248010635376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,16,2,balanced,0.0751091194152832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,16,2,balanced,0.07871744155883789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,16,2,balanced,0.07296895980834961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,16,2,balanced,0.08832256317138672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,16,2,balanced,0.09630335807800293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,16,2,balanced,0.12147583961486816
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,16,2,balanced,0.1480729579925537
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,16,2,balanced,0.22350208282470704
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,16,2,balanced,0.2683865547180176
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,16,2,balanced,0.3931903839111328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,16,2,balanced,0.5152473449707031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,16,2,power_law_1.01,0.05490687847137451
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,16,2,power_law_1.01,0.057208318710327145
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,16,2,balanced,0.7680473327636719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,16,2,power_law_1.01,0.06098944187164307
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,16,2,balanced,1.013785629272461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,16,2,power_law_1.01,0.11282303810119629
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,16,2,power_law_1.01,0.1118553638458252
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,16,2,power_law_1.01,0.11589887619018555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,16,2,power_law_1.01,0.11570560455322265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,16,2,power_law_1.01,0.11805312156677246
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,8,4,power_law_1.2,0.6224012756347657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,16,2,power_law_1.01,0.1183743953704834
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,16,2,power_law_1.01,0.12233087539672852
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,16,2,power_law_1.01,0.12410623550415038
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,16,2,power_law_1.01,0.13047552108764648
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,16,2,power_law_1.01,0.13712384223937987
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,16,2,power_law_1.01,0.1589798355102539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,16,2,power_law_1.01,0.14597375869750978
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,16,2,power_law_1.01,0.16742399215698242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,16,2,power_law_1.01,0.20685440063476562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,16,2,power_law_1.01,0.28973695755004886
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,16,2,power_law_1.01,0.11843199729919433
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,16,2,power_law_1.01,0.38135425567626957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,8,4,power_law_1.2,3.7337216186523436
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,16,2,power_law_1.01,0.6957901000976563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,16,2,power_law_1.01,0.5252454376220703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,16,2,power_law_1.01,1.3771827697753907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,16,2,power_law_1.01,0.03432960033416748
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,16,2,power_law_1.01,0.039233279228210446
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,16,2,power_law_1.01,0.04238207817077637
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,16,2,power_law_1.01,0.05746560096740723
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,16,2,power_law_1.01,0.05759871959686279
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,16,2,power_law_1.01,0.05842432022094727
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,16,2,power_law_1.01,0.05928959846496582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,16,2,power_law_1.01,0.06016895771026611
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,16,2,power_law_1.01,0.061030402183532714
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,16,2,power_law_1.01,0.9656575775146484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,16,2,power_law_1.01,0.06214399814605713
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,16,2,power_law_1.01,0.06351615905761718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,16,2,power_law_1.01,0.06688511848449707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,16,2,power_law_1.01,0.06884352207183839
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,16,2,power_law_1.01,0.07431807994842529
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,16,2,power_law_1.01,1.7882060241699218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,16,2,power_law_1.01,0.08040575981140137
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,16,2,power_law_1.01,0.0839590358734131
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,16,2,power_law_1.01,0.08536831855773926
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,16,2,power_law_1.01,0.09303808212280273
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,16,2,power_law_1.01,0.10108287811279297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,16,2,power_law_1.01,0.14714240074157714
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,16,2,power_law_1.01,0.1910041618347168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,16,2,power_law_1.01,0.2576832008361817
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,8,4,power_law_1.2,1.1084031677246093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,16,2,power_law_1.01,0.35025409698486326
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,8,4,power_law_1.2,4.937512817382812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,8,4,power_law_1.2,1.4274957275390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,16,2,power_law_1.2,0.05439487934112549
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,16,2,power_law_1.2,0.0609446382522583
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,16,2,power_law_1.2,0.11319040298461913
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,16,2,power_law_1.2,0.11217920303344726
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,16,2,power_law_1.01,0.5791462326049805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,16,2,power_law_1.2,0.11445631980895996
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,16,2,power_law_1.2,0.11559295654296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,16,2,power_law_1.2,0.11716095924377441
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,16,2,power_law_1.2,0.11820799827575683
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,16,2,power_law_1.2,0.06227200031280518
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,16,2,power_law_1.2,0.11978112220764162
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,16,2,power_law_1.2,0.12174079895019532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,16,2,power_law_1.01,2.6069375610351564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,16,2,power_law_1.2,0.12425984382629393
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,16,2,power_law_1.2,0.12977024078369143
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,16,2,power_law_1.2,0.13729151725769043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,16,2,power_law_1.2,0.15645952224731446
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,16,2,power_law_1.2,0.1480998420715332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,16,2,power_law_1.2,0.17783296585083008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,16,2,power_law_1.2,0.21345535278320313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,16,2,power_law_1.2,0.2906278419494629
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,16,2,power_law_1.01,3.5018200683593745
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,16,2,power_law_1.01,0.49012992858886717
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,16,2,power_law_1.2,0.3945996856689453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,16,2,power_law_1.2,0.5278860855102538
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,16,2,power_law_1.2,0.7210483551025391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,16,2,power_law_1.2,0.03370879888534546
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,16,2,power_law_1.2,0.039882240295410154
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,16,2,power_law_1.2,0.04123519897460938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,16,2,power_law_1.2,0.05702400207519531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,16,2,power_law_1.2,0.057349119186401364
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,16,2,power_law_1.2,0.05795839786529541
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,16,2,power_law_1.2,0.05881984233856201
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,16,2,power_law_1.2,0.05957632064819336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,16,2,power_law_1.2,0.060288000106811526
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,16,2,power_law_1.2,0.061040639877319336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,16,2,power_law_1.2,0.06341887950897217
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,16,2,power_law_1.2,0.0660211181640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,16,2,power_law_1.2,0.06830207824707031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,16,2,power_law_1.2,0.07318912029266358
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,16,2,power_law_1.2,1.8354150390625001
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,16,2,power_law_1.2,0.08609279632568359
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,16,2,power_law_1.2,0.9480473327636719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,16,2,power_law_1.2,0.08065279960632324
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,16,2,power_law_1.2,0.081529598236084
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,16,2,power_law_1.2,0.09310591697692872
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,16,2,power_law_1.2,0.10213248252868652
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,16,2,power_law_1.01,0.9142591857910156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,16,2,power_law_1.2,0.14615936279296876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,16,2,power_law_1.2,0.1890713691711426
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,16,2,power_law_1.2,1.2907186889648439
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,16,2,power_law_1.2,0.2595136070251465
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,16,2,power_law_1.01,1.1810546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,32,1,balanced,0.045841917991638184
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,16,2,power_law_1.2,0.3454476928710938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,32,1,balanced,0.062245759963989254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,32,1,balanced,0.07248064041137695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,32,1,balanced,0.0737779188156128
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,32,1,balanced,0.11224448204040527
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,32,1,balanced,0.11450559616088866
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,32,1,balanced,0.11634495735168457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,32,1,balanced,0.11790207862854003
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,32,1,balanced,0.11875007629394532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,32,1,balanced,0.12019200325012207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,32,1,balanced,0.12209535598754884
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,32,1,balanced,0.12495807647705078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,32,1,balanced,0.1269337558746338
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,32,1,balanced,0.13289279937744142
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,32,1,balanced,0.14289152145385742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,32,1,balanced,0.14631872177124022
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,32,1,balanced,0.144846715927124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,32,1,balanced,0.16648128509521484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,32,1,balanced,0.18273023605346678
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,32,1,balanced,0.2427456092834473
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,32,1,balanced,0.3362854385375976
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,32,1,balanced,0.48442047119140624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,32,1,balanced,0.641415023803711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,32,1,balanced,0.9471481323242188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,32,1,balanced,1.253219223022461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,16,2,power_law_1.2,0.5769036865234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,32,1,balanced,1.87808837890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,32,1,balanced,0.0331987190246582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,32,1,balanced,0.04010176181793213
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,32,1,balanced,0.05293632030487061
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,32,1,balanced,0.05355008125305176
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,32,1,balanced,0.06672639846801758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,32,1,balanced,0.06738431930541992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,32,1,balanced,0.06863679885864257
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,32,1,balanced,0.0689798402786255
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,32,1,balanced,0.06954048156738282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,32,1,balanced,0.06977663993835449
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,32,1,balanced,0.0713747215270996
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,32,1,balanced,0.07235775947570801
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,32,1,balanced,0.0741971206665039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,32,1,balanced,0.07671807765960693
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,32,1,balanced,2.4899200439453124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,32,1,balanced,0.07924032211303711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,32,1,balanced,0.08182527542114257
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,32,1,balanced,0.08550399780273438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,32,1,balanced,0.09393407821655272
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,32,1,balanced,0.10077376365661621
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,32,1,balanced,0.13559103965759278
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,32,1,balanced,0.15285696029663084
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,32,1,balanced,0.2360095977783203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,16,2,power_law_1.2,2.690205383300781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,32,1,balanced,0.2864761543273926
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,32,1,balanced,0.41145217895507813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,16,2,power_law_1.2,0.48327297210693365
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,32,1,power_law_1.01,0.05494719982147217
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,32,1,balanced,0.53853759765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,32,1,power_law_1.01,0.06466432094573975
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,32,1,power_law_1.01,0.07324096202850341
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,32,1,balanced,0.7967021179199218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,32,1,power_law_1.01,0.11438207626342774
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,32,1,balanced,1.041212158203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,32,1,power_law_1.01,0.11661439895629884
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,32,1,power_law_1.01,0.1173523235321045
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,32,1,power_law_1.01,0.11909312248229979
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,32,1,power_law_1.01,0.12100223541259765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,32,1,power_law_1.01,0.06063807964324951
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,32,1,power_law_1.01,0.12397952079772949
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,32,1,power_law_1.01,0.12592703819274903
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,32,1,power_law_1.01,0.11170751571655273
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,32,1,power_law_1.01,0.1387116813659668
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,32,1,power_law_1.01,0.14283647537231445
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,32,1,power_law_1.01,0.1463167953491211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,32,1,power_law_1.01,0.15071616172790528
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,32,1,power_law_1.01,0.17114688873291015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,32,1,power_law_1.01,0.13057408332824708
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,32,1,power_law_1.01,0.1931167984008789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,32,1,power_law_1.01,0.27812416076660157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,32,1,power_law_1.01,0.36037952423095704
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,32,1,power_law_1.01,0.5049817657470703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,32,1,power_law_1.01,0.6600070190429688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,32,1,power_law_1.01,0.035317120552062986
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,32,1,power_law_1.01,0.04184192180633545
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,32,1,power_law_1.01,0.04406784057617187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,32,1,power_law_1.01,0.052945280075073244
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,32,1,power_law_1.01,0.06666175842285156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,32,1,power_law_1.01,0.06729023933410645
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,32,1,power_law_1.01,0.06823808193206787
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,32,1,power_law_1.01,0.0690944004058838
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,32,1,power_law_1.01,0.06948544025421143
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,32,1,power_law_1.01,0.07020864009857178
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,32,1,power_law_1.01,0.0718713617324829
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,32,1,power_law_1.01,0.9567884826660157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,32,1,power_law_1.01,0.07322495937347412
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,32,1,power_law_1.01,0.07645567893981933
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,32,1,power_law_1.01,0.07681983947753906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,16,2,power_law_1.2,0.9690815734863282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,32,1,power_law_1.01,0.0797382402420044
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,32,1,power_law_1.01,0.08424320220947265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,32,1,power_law_1.01,0.08728384017944336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,32,1,power_law_1.01,0.09887167930603027
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,16,2,power_law_1.2,3.380044860839844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,32,1,power_law_1.01,0.10644096374511718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,32,1,power_law_1.01,0.13518912315368653
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,32,1,power_law_1.01,0.1912825584411621
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,32,1,power_law_1.01,1.263545608520508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,32,1,power_law_1.01,0.24057024002075195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,32,1,power_law_1.01,0.3203398513793945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,32,1,power_law_1.2,0.05484864234924316
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,32,1,power_law_1.2,0.06059648036956787
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,32,1,power_law_1.2,0.06256319999694823
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,32,1,power_law_1.2,0.07290112018585206
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,32,1,power_law_1.2,0.1110854434967041
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,32,1,power_law_1.2,0.1145043182373047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,32,1,power_law_1.2,0.11668416023254395
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,32,1,power_law_1.2,0.11732031822204589
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,32,1,power_law_1.2,0.1197548770904541
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,32,1,power_law_1.01,0.44803199768066404
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,32,1,power_law_1.2,0.12131584167480469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,32,1,power_law_1.2,0.12295871734619142
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,32,1,power_law_1.2,0.12585920333862305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,32,1,power_law_1.2,0.1294003200531006
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,32,1,power_law_1.2,0.138536958694458
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,32,1,power_law_1.2,0.14235456466674806
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,32,1,power_law_1.2,0.14690303802490234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,32,1,power_law_1.2,0.14972031593322754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,32,1,power_law_1.01,1.883338165283203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,32,1,power_law_1.2,0.17137664794921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,32,1,power_law_1.2,0.19429759979248046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,32,1,power_law_1.01,0.5472927856445312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,32,1,power_law_1.2,0.27873600006103516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,32,1,power_law_1.2,0.36151168823242186
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,32,1,power_law_1.01,2.4876832580566406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,32,1,power_law_1.2,0.5015558242797852
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,32,1,power_law_1.01,0.8156179046630859
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,32,1,power_law_1.2,0.03734143972396851
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,32,1,power_law_1.2,0.04222911834716797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,32,1,power_law_1.2,0.04309760093688965
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,32,1,power_law_1.2,0.052990078926086426
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,32,1,power_law_1.2,0.06649856090545654
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,32,1,power_law_1.2,0.06768383979797363
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,32,1,power_law_1.2,0.06842751979827881
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,32,1,power_law_1.2,0.06928832054138183
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,32,1,power_law_1.2,0.07014976024627687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,32,1,power_law_1.2,0.6611436462402344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,32,1,power_law_1.2,0.07079103946685791
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,32,1,power_law_1.2,0.07151231765747071
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,32,1,power_law_1.2,0.07395455837249756
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,32,1,power_law_1.2,0.07453824043273925
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,32,1,power_law_1.2,0.07756159782409669
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,32,1,power_law_1.2,0.07958784103393554
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,32,1,power_law_1.2,0.08483455657958985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,32,1,power_law_1.2,0.08532671928405762
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,32,1,power_law_1.2,0.09765824317932129
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,16,2,power_law_1.2,1.2582681274414063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,32,1,power_law_1.01,1.0523673248291014
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,32,1,power_law_1.2,0.10587200164794922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,32,1,power_law_1.2,0.1336844825744629
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,32,1,power_law_1.2,0.9572077178955078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,32,1,power_law_1.2,0.1912691116333008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,32,1,power_law_1.2,0.23924480438232423
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,32,1,power_law_1.2,0.31994176864624024
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,32,1,power_law_1.2,1.26474365234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,1,32,balanced,0.050245118141174314
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,1,32,balanced,0.05026944160461426
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,32,1,power_law_1.2,0.449224967956543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,1,32,balanced,0.05034239768981934
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,1,32,balanced,0.06815616130828858
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,1,32,balanced,0.19347967147827147
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,1,32,balanced,0.10072447776794433
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,32,1,power_law_1.2,0.5532364654541015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,32,1,power_law_1.2,1.8761138916015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,1,32,balanced,0.20639488220214847
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,1,32,balanced,0.1996518325805664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,1,32,balanced,0.18440704345703124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,1,32,balanced,0.18428287506103516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,1,32,balanced,0.20945152282714843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,1,32,balanced,0.2116428756713867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,1,32,balanced,0.2095244789123535
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,1,32,balanced,0.21212287902832033
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,32,1,power_law_1.2,0.7987232208251953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,1,32,balanced,0.2147212791442871
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,32,1,power_law_1.2,2.4824896240234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,1,32,balanced,0.2180044746398926
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,1,32,balanced,0.2118694305419922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,1,32,balanced,0.2205734443664551
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,1,32,balanced,0.23406208038330076
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,1,32,balanced,0.2523891258239746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,1,32,balanced,0.2606399917602539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,1,32,balanced,0.9112640380859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,32,1,power_law_1.2,1.0522898864746093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,1,32,balanced,0.28955007553100587
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,1,32,balanced,0.32514942169189454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,1,32,balanced,0.4405503845214843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,1,32,balanced,0.5068352127075195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,1,32,balanced,0.06491903781890869
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,1,32,balanced,0.7011033630371093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,1,32,balanced,0.03954943895339966
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,1,32,balanced,0.03955456018447876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,1,32,balanced,0.112739839553833
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,1,32,balanced,0.1146828842163086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,1,32,balanced,0.11749376296997069
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,1,32,balanced,0.039267840385437014
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,1,32,balanced,0.11705727577209472
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,1,32,balanced,0.1133631992340088
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,1,32,balanced,0.051891198158264165
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,1,32,balanced,0.1316096019744873
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,1,32,balanced,0.14637951850891112
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,1,32,balanced,0.16146560668945314
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,1,32,balanced,0.18118272781372072
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,1,32,balanced,0.11807744026184082
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,1,32,balanced,0.11582592010498047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,1,32,balanced,0.12080512046813965
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,1,32,balanced,0.21227392196655273
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,1,32,balanced,0.24584831237792967
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,1,32,balanced,0.7104370880126953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,1,32,balanced,0.12782464027404786
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,1,32,power_law_1.01,0.1854092788696289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,1,32,balanced,0.13309823989868164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,1,32,power_law_1.01,0.18998783111572265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,1,32,balanced,0.12561535835266113
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,1,32,balanced,0.13408767700195312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,1,32,power_law_1.01,0.19947904586791992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,1,32,power_law_1.01,0.18097280502319335
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,1,32,power_law_1.01,0.17987327575683593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,1,32,balanced,0.33420543670654296
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,1,32,balanced,0.40089599609375004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,1,32,balanced,0.5506201553344726
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,1,32,power_law_1.01,0.2055129623413086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,1,32,power_law_1.01,0.18213760375976562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,1,32,power_law_1.01,0.16896255493164064
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,1,32,power_law_1.01,0.14583423614501953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,1,32,power_law_1.01,0.20839040756225585
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,1,32,power_law_1.01,0.18610431671142577
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,1,32,power_law_1.01,0.2111846351623535
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,1,32,power_law_1.01,0.19682687759399414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,1,32,power_law_1.01,0.19809919357299804
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,1,32,power_law_1.01,0.2159270477294922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,1,32,power_law_1.01,0.21985408782958987
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,1,32,power_law_1.01,0.8987264251708984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,1,32,power_law_1.01,1.2304499053955078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,1,32,power_law_1.01,0.3033036804199219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,1,32,power_law_1.01,0.3348147201538086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,1,32,power_law_1.01,0.1153331184387207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,1,32,power_law_1.01,0.3884761428833008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,1,32,power_law_1.01,1.7878297424316407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,1,32,power_law_1.01,2.253955841064453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,1,32,power_law_1.01,0.1115187168121338
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,1,32,power_law_1.01,0.2156889533996582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,1,32,power_law_1.01,0.5453670501708985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,1,32,power_law_1.01,0.10592384338378906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,1,32,power_law_1.01,0.25423744201660153
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,1,32,power_law_1.01,0.08576512336730957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,1,32,power_law_1.01,0.11708288192749024
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,1,32,power_law_1.01,0.09999360084533691
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,1,32,power_law_1.01,0.10042367935180666
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,1,32,power_law_1.01,0.10783616065979004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,1,32,power_law_1.01,0.10806655883789062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,1,32,power_law_1.01,0.10859264373779295
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,1,32,power_law_1.01,0.1412928009033203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,1,32,power_law_1.01,0.1585574436187744
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,1,32,power_law_1.01,0.10616703987121583
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,1,32,power_law_1.01,0.19423616409301758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,1,32,power_law_1.01,0.23519744873046874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,1,32,power_law_1.01,0.2670028877258301
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,1,32,power_law_1.01,0.12209152221679687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,1,32,power_law_1.01,0.5847462463378906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,1,32,power_law_1.01,0.136495361328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,1,32,power_law_1.01,0.7151961517333985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,1,32,power_law_1.01,0.8829798126220704
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,1,32,power_law_1.01,0.13353216171264648
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,1,32,power_law_1.2,0.18536191940307617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,1,32,power_law_1.2,0.20029056549072263
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,1,32,power_law_1.01,1.2891366577148438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,1,32,power_law_1.01,1.806544647216797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,1,32,power_law_1.01,0.12662015914916994
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,1,32,power_law_1.2,0.17999616622924805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,1,32,power_law_1.01,0.1279539203643799
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,1,32,power_law_1.2,0.16038015365600586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,1,32,power_law_1.2,0.18042495727539062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,1,32,power_law_1.2,0.1937241554260254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,1,32,power_law_1.2,0.1795020866394043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,1,32,power_law_1.2,0.17747968673706055
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,1,32,power_law_1.2,0.18836223602294921
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,1,32,power_law_1.2,0.20257280349731444
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,1,32,power_law_1.2,0.1491430377960205
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,1,32,power_law_1.2,0.20486656188964844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,1,32,power_law_1.2,0.21884416580200194
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,1,32,power_law_1.2,0.2982694435119629
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,1,32,power_law_1.2,0.333573112487793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,1,32,power_law_1.2,0.3793907165527344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,1,32,power_law_1.2,0.4752550506591797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,1,32,power_law_1.2,0.17551488876342775
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,1,32,power_law_1.2,0.17909120559692382
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,1,32,power_law_1.2,1.2065164947509766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,1,32,power_law_1.2,0.17322751998901367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,1,32,power_law_1.01,0.3961484909057617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,1,32,power_law_1.01,0.3602751922607422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,1,32,power_law_1.2,1.422592010498047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,1,32,power_law_1.2,2.4498419189453124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,1,32,power_law_1.2,0.11751680374145508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,1,32,power_law_1.2,0.09429632186889648
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,1,32,power_law_1.2,3.077236328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,1,32,power_law_1.2,0.08816767692565917
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,1,32,power_law_1.2,0.11063551902770996
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,1,32,power_law_1.2,0.11248512268066406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,1,32,power_law_1.2,0.10986111640930177
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,1,32,power_law_1.2,0.11681920051574708
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,1,32,power_law_1.2,0.12632960319519043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,1,32,power_law_1.2,0.10547200202941895
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,1,32,power_law_1.2,0.20891647338867186
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,1,32,power_law_1.2,0.1288576030731201
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,1,32,power_law_1.2,0.10505215644836426
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,1,32,power_law_1.2,0.11498111724853514
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,1,32,power_law_1.2,0.18313087463378905
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,1,32,power_law_1.2,0.2229862403869629
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,1,32,power_law_1.2,0.12907520294189453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,1,32,power_law_1.2,0.25941631317138675
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,1,32,power_law_1.2,0.41886463165283205
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,1,32,power_law_1.2,0.10815103530883789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,1,32,power_law_1.2,0.8547020721435548
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,1,32,power_law_1.2,1.0033561706542968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,1,32,power_law_1.2,0.7955084991455078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,1,32,power_law_1.2,0.631701774597168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,1,32,power_law_1.2,0.13033087730407716
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,1,32,power_law_1.2,1.6307916259765627
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,1,32,power_law_1.2,2.1658444213867187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,1,32,power_law_1.2,0.12775039672851562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,1,32,power_law_1.2,0.14343808174133302
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,2,16,balanced,0.03128704071044922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,2,16,balanced,0.03130111932754517
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,2,16,balanced,0.045355520248413085
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,2,16,balanced,0.06872447967529297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,2,16,balanced,0.10489088058471681
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,2,16,balanced,0.189617919921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,2,16,balanced,0.195865592956543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,2,16,balanced,0.2130995178222656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,2,16,balanced,0.19050880432128908
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,2,16,balanced,0.19484415054321289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,2,16,balanced,0.19582080841064453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,2,16,balanced,0.20092287063598632
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,2,16,balanced,0.20158336639404295
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,2,16,balanced,0.2424550437927246
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,2,16,balanced,0.26620031356811524
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,2,16,balanced,0.2894976043701172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,2,16,balanced,0.34276222229003905
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,2,16,balanced,0.19814783096313476
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,2,16,balanced,0.20847616195678711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,2,16,balanced,0.5456703948974609
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,2,16,balanced,0.21058176040649412
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,2,16,balanced,0.6506687927246093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,1,32,power_law_1.2,0.3104499244689941
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,2,16,balanced,0.03285376071929932
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,2,16,balanced,0.032700159549713136
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,2,16,balanced,0.039205119609832764
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,2,16,balanced,0.048463358879089355
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,2,16,balanced,0.06748415946960448
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,2,16,balanced,0.9505216217041015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,2,16,balanced,0.11159423828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,2,16,balanced,0.12303232192993163
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,2,16,balanced,0.12946559906005858
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,2,16,balanced,1.2425023651123046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,2,16,balanced,0.12594688415527344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,2,16,balanced,0.12558079719543458
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,2,16,balanced,0.12931712150573732
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,1,32,power_law_1.2,0.5360460662841797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,2,16,balanced,0.13196800231933595
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,2,16,balanced,0.1292518424987793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,2,16,balanced,0.13225855827331542
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,2,16,balanced,0.13619711875915527
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,2,16,balanced,0.14213248252868654
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,2,16,balanced,0.1555123233795166
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,2,16,balanced,0.16950016021728515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,2,16,balanced,0.19738496780395506
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,2,16,balanced,0.22188928604125974
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,2,16,balanced,0.27042303085327146
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,2,16,balanced,0.32022270202636716
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,2,16,balanced,0.13471360206604005
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,2,16,balanced,0.454856948852539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,2,16,balanced,0.5565875244140626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,2,16,power_law_1.01,0.09990655899047851
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,2,16,balanced,0.21099647521972656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,2,16,balanced,0.7917350769042969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,2,16,balanced,0.228460807800293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,2,16,balanced,1.0350233459472657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,2,16,power_law_1.01,0.17879295349121094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,2,16,balanced,0.40152832031249996
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,2,16,power_law_1.01,0.18372095108032227
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,2,16,power_law_1.01,0.18387327194213868
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,2,16,power_law_1.01,0.1818623924255371
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,2,16,power_law_1.01,0.17649919509887696
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,2,16,power_law_1.01,0.17351295471191405
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,2,16,power_law_1.01,0.17414655685424804
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,2,16,power_law_1.01,0.17750272750854493
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,2,16,power_law_1.01,0.19028480529785158
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,2,16,power_law_1.01,0.20762496948242187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,2,16,power_law_1.01,0.1784832000732422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,2,16,power_law_1.01,0.21581184387207034
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,2,16,power_law_1.01,0.25955583572387697
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,2,16,power_law_1.01,0.2702233505249024
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,2,16,power_law_1.01,0.49745918273925777
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,2,16,power_law_1.01,0.3614438247680664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,2,16,power_law_1.01,0.40880256652832025
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,2,16,power_law_1.01,0.2253209686279297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,2,16,power_law_1.01,0.21323776245117188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,2,16,power_law_1.01,0.06596864223480224
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,2,16,power_law_1.01,0.10796544075012207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,2,16,power_law_1.01,0.10851967811584473
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,2,16,power_law_1.01,0.10718208312988282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,2,16,power_law_1.01,0.2311257553100586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,2,16,power_law_1.01,0.10428159713745118
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,2,16,power_law_1.01,0.11395199775695801
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,2,16,power_law_1.01,0.2509708786010742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,2,16,power_law_1.01,0.11205120086669922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,2,16,power_law_1.01,0.10800512313842774
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,2,16,power_law_1.01,0.11260160446166992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,2,16,power_law_1.01,0.11876607894897462
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,2,16,power_law_1.01,0.11775487899780272
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,2,16,power_law_1.01,0.1213798427581787
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,2,16,power_law_1.01,0.1263308811187744
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,2,16,power_law_1.01,0.12727807998657226
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,2,16,power_law_1.01,0.12547072410583496
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,2,16,power_law_1.01,0.12841471672058105
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,2,16,power_law_1.01,0.14228992462158202
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,2,16,power_law_1.01,0.17386367797851562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,2,16,power_law_1.01,0.20712959289550784
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,2,16,power_law_1.01,0.24564992904663088
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,2,16,power_law_1.01,0.26952320098876953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,2,16,power_law_1.01,0.3649766540527344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,2,16,power_law_1.01,0.4604940795898438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,2,16,power_law_1.01,0.6684134674072266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,2,16,power_law_1.01,0.7800832366943359
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,2,16,power_law_1.01,1.3092941284179687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,2,16,power_law_1.01,1.616779479980469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,2,16,power_law_1.01,0.6590886688232421
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,2,16,power_law_1.01,0.904136962890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,2,16,power_law_1.01,1.0982988739013673
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,2,16,power_law_1.2,0.0995263957977295
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,2,16,power_law_1.01,1.7689587402343752
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,2,16,power_law_1.01,2.3629977416992185
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,2,16,power_law_1.2,0.1790412712097168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,2,16,power_law_1.2,0.18000511169433592
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,2,16,power_law_1.2,0.16239999771118163
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,2,16,power_law_1.2,0.17981952667236328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,2,16,power_law_1.2,0.16838016510009765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,2,16,power_law_1.2,0.17393152236938475
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,2,16,power_law_1.2,0.18840063095092774
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,2,16,power_law_1.2,0.23108224868774413
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,2,16,power_law_1.2,0.1993804740905762
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,2,16,power_law_1.2,0.22698240280151366
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,2,16,power_law_1.2,0.20971136093139647
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,2,16,power_law_1.2,0.18327808380126953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,2,16,power_law_1.2,0.19085567474365234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,2,16,power_law_1.2,0.42929664611816404
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,2,16,power_law_1.2,0.28034048080444335
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,2,16,power_law_1.2,0.6438182067871094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,2,16,power_law_1.2,1.0492658996582032
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,2,16,power_law_1.2,1.285400390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,2,16,power_law_1.2,1.8772915649414064
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,2,16,power_law_1.2,0.07061247825622559
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,2,16,power_law_1.2,0.10804224014282227
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,2,16,power_law_1.2,0.22419328689575196
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,2,16,power_law_1.2,0.09656959533691406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,2,16,power_law_1.2,0.10728575706481933
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,2,16,power_law_1.2,2.441297912597656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,2,16,power_law_1.2,0.10205183982849123
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,2,16,power_law_1.2,0.10078335762023927
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,2,16,power_law_1.2,0.11262592315673828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,2,16,power_law_1.2,0.23761920928955077
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,2,16,power_law_1.2,0.40335742950439457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,2,16,power_law_1.2,0.10977536201477052
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,2,16,power_law_1.2,0.11139328002929687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,2,16,power_law_1.2,0.3328051376342773
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,2,16,power_law_1.2,0.1139673614501953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,2,16,power_law_1.2,0.12230655670166016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,2,16,power_law_1.2,0.12868608474731444
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,2,16,power_law_1.2,0.1239577579498291
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,2,16,power_law_1.2,0.13018752098083497
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,2,16,power_law_1.2,0.13573887825012207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,2,16,power_law_1.2,0.1444108772277832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,2,16,power_law_1.2,0.19007104873657227
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,2,16,power_law_1.2,0.2186534309387207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,2,16,power_law_1.2,0.2697561645507812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,2,16,power_law_1.2,0.30818815231323243
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,2,16,power_law_1.2,0.37374977111816404
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,2,16,power_law_1.2,0.5556108856201172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,2,16,power_law_1.2,0.7288537597656249
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,4,8,balanced,0.03192447900772095
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,2,16,power_law_1.2,0.9515289306640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,4,8,balanced,0.05514624118804932
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,2,16,power_law_1.2,0.21716096878051755
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,4,8,balanced,0.06349311828613281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,4,8,balanced,0.08014080047607422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,2,16,power_law_1.2,0.12291711807250975
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,4,8,balanced,0.11274880409240721
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,4,8,balanced,0.18543487548828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,4,8,balanced,0.24788991928100587
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,2,16,power_law_1.2,1.3888627624511718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,4,8,balanced,0.2532454490661621
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,4,8,balanced,0.2539430427551269
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,4,8,balanced,0.2593024063110351
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,4,8,balanced,0.25107072830200194
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,4,8,balanced,0.2672102355957031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,4,8,balanced,0.25811967849731443
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,4,8,balanced,0.2797439956665039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,4,8,balanced,0.2879334449768066
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,2,16,power_law_1.2,1.9337957763671874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,4,8,balanced,0.3170969581604004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,4,8,balanced,0.3520780944824219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,4,8,balanced,0.40707839965820314
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,4,8,balanced,0.4606131362915039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,4,8,balanced,0.2548044776916504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,4,8,balanced,0.6579353332519531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,4,8,balanced,0.7806924438476562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,4,8,balanced,0.2734182357788086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,4,8,balanced,0.032325119972229005
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,4,8,balanced,0.0334553599357605
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,4,8,balanced,0.0418943977355957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,2,16,power_law_1.2,0.5555353546142578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,4,8,balanced,1.0763289642333984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,4,8,balanced,0.05284224033355713
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,4,8,balanced,0.07518335819244384
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,4,8,balanced,0.11919743537902831
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,4,8,balanced,0.13574015617370605
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,4,8,balanced,0.13913599967956541
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,4,8,balanced,0.1415283203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,4,8,balanced,0.14462335586547853
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,4,8,balanced,1.3909056091308594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,4,8,balanced,0.13871871948242187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,4,8,balanced,0.14179327964782715
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,4,8,balanced,0.14824447631835938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,4,8,balanced,0.15433088302612305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,4,8,balanced,0.16036224365234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,4,8,balanced,0.1729203224182129
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,4,8,balanced,0.19523839950561522
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,4,8,balanced,0.2187775993347168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,4,8,balanced,0.2633305549621582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,4,8,balanced,2.00695556640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,4,8,balanced,0.306246395111084
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,4,8,balanced,0.3953190231323242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,4,8,balanced,0.486187515258789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,4,8,power_law_1.01,0.07930496215820312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,4,8,balanced,0.1360204792022705
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,4,8,balanced,0.7098534393310547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,4,8,power_law_1.01,0.10861951828002929
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,4,8,power_law_1.01,0.13118335723876953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,4,8,balanced,0.8835929870605469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,4,8,balanced,2.585113525390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,4,8,power_law_1.01,0.13565695762634278
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,4,8,power_law_1.01,0.16101760864257814
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,4,8,power_law_1.01,0.1644326400756836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,4,8,power_law_1.01,0.2373593521118164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,4,8,power_law_1.01,0.2237183952331543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,4,8,power_law_1.01,0.2377497673034668
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,4,8,power_law_1.01,0.22163455963134765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,4,8,power_law_1.01,0.23216512680053708
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,4,8,power_law_1.01,0.23964799880981444
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,4,8,power_law_1.01,0.23588863372802735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,4,8,power_law_1.01,0.24483327865600585
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,4,8,power_law_1.01,0.24783744812011718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,4,8,power_law_1.01,0.25180160522460937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,4,8,balanced,1.6915277099609376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,4,8,power_law_1.01,0.31723775863647463
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,4,8,power_law_1.01,0.34895233154296873
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,4,8,power_law_1.01,0.46334209442138674
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,4,8,power_law_1.01,0.508666877746582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,4,8,power_law_1.01,0.7043238067626953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,4,8,balanced,1.2853721618652343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,4,8,power_law_1.01,0.8591180419921874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,4,8,power_law_1.01,0.051024641990661625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,4,8,power_law_1.01,0.27250560760498044
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,4,8,power_law_1.01,0.07033472061157227
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,4,8,power_law_1.01,1.2508787536621093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,4,8,power_law_1.01,0.08432127952575683
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,4,8,power_law_1.01,0.08642560005187988
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,4,8,power_law_1.01,0.10514687538146972
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,4,8,power_law_1.01,0.10710528373718262
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,4,8,power_law_1.01,0.11843328475952149
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,4,8,power_law_1.01,1.5506483459472657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,4,8,power_law_1.01,0.12550911903381348
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,4,8,power_law_1.01,0.1289971160888672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,4,8,power_law_1.01,0.1329638385772705
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,4,8,power_law_1.01,0.13202048301696778
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,4,8,power_law_1.01,0.13757568359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,4,8,power_law_1.01,0.13993215560913086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,4,8,power_law_1.01,2.2439167785644534
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,4,8,power_law_1.01,0.14381567955017088
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,4,8,power_law_1.01,0.15970687866210936
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,4,8,power_law_1.01,0.17188352584838867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,4,8,power_law_1.01,0.21841663360595703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,4,8,power_law_1.01,0.26966272354125975
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,4,8,power_law_1.01,0.1226585578918457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,4,8,power_law_1.01,0.315866870880127
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,4,8,power_law_1.01,0.38447616577148436
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,4,8,power_law_1.01,3.2152090454101563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,4,8,power_law_1.01,0.4856217575073242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,4,8,power_law_1.2,0.08004863739013672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,4,8,power_law_1.2,0.11451775550842284
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,4,8,power_law_1.01,0.6779264068603515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,4,8,power_law_1.01,0.11848575592041015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,4,8,power_law_1.2,0.12798975944519042
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,4,8,power_law_1.2,0.15545599937438964
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,4,8,power_law_1.2,0.16564096450805663
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,4,8,power_law_1.2,0.2191756820678711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,4,8,power_law_1.2,0.23973503112792968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,4,8,power_law_1.2,0.22914815902709962
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,4,8,power_law_1.2,0.22257280349731445
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,4,8,power_law_1.2,0.23895296096801757
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,4,8,power_law_1.2,0.22544384002685547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,4,8,power_law_1.2,0.24082176208496095
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,4,8,power_law_1.2,0.234704647064209
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,4,8,power_law_1.2,0.2459212875366211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,4,8,power_law_1.01,1.7238336181640626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,4,8,power_law_1.2,0.25892351150512694
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,4,8,power_law_1.2,0.2744332885742188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,4,8,power_law_1.2,0.34173439025878904
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,4,8,power_law_1.2,0.3891136169433594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,4,8,power_law_1.2,0.46481662750244135
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,4,8,power_law_1.01,1.3139942932128907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,4,8,power_law_1.2,0.5664038467407226
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,4,8,power_law_1.2,0.10691200256347658
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,4,8,power_law_1.01,0.8209945678710937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,4,8,power_law_1.2,0.9472959899902345
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,4,8,power_law_1.2,0.05164031982421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,4,8,power_law_1.2,0.0704038381576538
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,4,8,power_law_1.2,0.076495361328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,4,8,power_law_1.2,0.08414463996887207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,4,8,power_law_1.2,1.3593177795410156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,4,8,power_law_1.2,0.10507264137268066
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,4,8,power_law_1.2,0.1071513557434082
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,4,8,power_law_1.2,0.11691904067993164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,4,8,power_law_1.2,0.12964863777160646
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,4,8,power_law_1.2,0.7841126251220703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,4,8,power_law_1.2,0.11811200141906739
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,4,8,power_law_1.2,1.768536376953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,4,8,power_law_1.2,0.12499072074890136
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,4,8,power_law_1.2,0.12508928298950195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,4,8,power_law_1.2,0.12763775825500487
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,4,8,power_law_1.2,0.13232383728027344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,4,8,power_law_1.2,0.13693183898925781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,4,8,power_law_1.2,0.1437452793121338
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,4,8,power_law_1.2,0.16112127304077148
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,4,8,power_law_1.2,0.20535295486450195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,4,8,power_law_1.2,0.22074752807617187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,4,8,power_law_1.2,2.52621826171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,4,8,power_law_1.2,0.2658803176879883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,4,8,power_law_1.2,0.3155583953857422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,4,8,power_law_1.2,0.427696647644043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,4,8,power_law_1.2,0.5292812728881836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,8,4,balanced,0.03256063938140869
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,8,4,balanced,0.055371518135070796
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,8,4,balanced,0.06434944152832031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,4,8,power_law_1.2,0.13109631538391114
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,8,4,balanced,0.08266624450683593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,4,8,power_law_1.2,0.801312026977539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,8,4,balanced,0.11186047554016114
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,8,4,balanced,0.18248064041137696
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,4,8,power_law_1.2,0.9619967651367187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,4,8,power_law_1.2,3.726007080078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,8,4,balanced,0.18500864028930664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,8,4,balanced,0.18755968093872072
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,8,4,balanced,0.24147455215454103
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,8,4,balanced,0.24235519409179687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,8,4,balanced,0.24484479904174802
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,8,4,balanced,0.2479360008239746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,8,4,balanced,0.2574835205078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,8,4,balanced,0.26329216003417966
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,8,4,balanced,0.26739072799682617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,8,4,balanced,0.27498239517211914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,8,4,balanced,0.2968819236755371
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,8,4,balanced,0.32082687377929686
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,8,4,balanced,0.3765209579467773
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,4,8,power_law_1.2,1.9049600219726563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,8,4,balanced,0.4245606231689453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,8,4,balanced,0.643983383178711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,8,4,balanced,0.2525708770751953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,4,8,power_law_1.2,1.3463859558105469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,8,4,balanced,0.7912550354003907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,8,4,balanced,0.031000320911407468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,8,4,balanced,0.03290112018585205
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,8,4,balanced,0.04294400215148926
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,8,4,balanced,1.0856985473632812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,8,4,balanced,0.07917183876037598
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,8,4,balanced,0.1318835163116455
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,8,4,balanced,0.13371904373168947
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,8,4,balanced,0.13515007972717286
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,8,4,balanced,1.403618621826172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,8,4,balanced,0.16025856018066406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,8,4,balanced,0.1633113670349121
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,8,4,balanced,0.1650048065185547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,8,4,balanced,0.16783231735229492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,8,4,balanced,0.1732761573791504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,8,4,balanced,0.17787136077880858
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,8,4,balanced,1.984075469970703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,8,4,balanced,0.1826278305053711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,8,4,balanced,0.05861504077911377
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,8,4,balanced,0.19201919555664063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,8,4,balanced,0.21208959579467773
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,8,4,balanced,0.23420288085937502
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,8,4,balanced,0.27694847106933596
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,8,4,balanced,0.3166067123413086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,8,4,balanced,0.39989376068115234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,8,4,balanced,0.48483455657958985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,8,4,power_law_1.01,0.06399615764617919
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,8,4,power_law_1.01,0.07654016017913819
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,8,4,balanced,0.7308953857421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,8,4,power_law_1.01,0.09034496307373047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,8,4,balanced,0.8978508758544923
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,8,4,power_law_1.01,0.10408320426940917
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,8,4,power_law_1.01,0.11936256408691406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,8,4,power_law_1.01,0.1364083194732666
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,8,4,balanced,0.1585753631591797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,8,4,power_law_1.01,0.16694143295288086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,8,4,power_law_1.01,0.16700927734374998
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,8,4,balanced,1.3105587768554687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,8,4,power_law_1.01,0.21329023361206056
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,8,4,balanced,2.5847207641601564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,8,4,power_law_1.01,0.22346368789672852
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,8,4,power_law_1.01,0.22022655487060544
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,8,4,power_law_1.01,0.23458688735961913
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,8,4,power_law_1.01,0.2305062484741211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,8,4,power_law_1.01,0.2397273635864258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,8,4,balanced,1.7190797424316409
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,8,4,power_law_1.01,0.2628595161437988
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,8,4,power_law_1.01,0.29558912277221683
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,8,4,power_law_1.01,0.3480051040649414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,8,4,power_law_1.01,0.427507209777832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,8,4,power_law_1.01,0.5211532974243165
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,8,4,power_law_1.01,0.7003238677978516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,8,4,power_law_1.01,0.8425011444091798
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,8,4,power_law_1.01,0.24301567077636718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,8,4,power_law_1.01,0.041152000427246094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,8,4,power_law_1.01,0.24537343978881837
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,8,4,power_law_1.01,0.05453184127807618
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,8,4,power_law_1.01,1.1771788787841797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,8,4,power_law_1.01,0.06478079795837402
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,8,4,power_law_1.01,0.07293824195861817
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,8,4,power_law_1.01,0.0913049602508545
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,8,4,power_law_1.01,0.09665023803710937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,8,4,power_law_1.01,1.4728498840332032
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,8,4,power_law_1.01,0.12235775947570801
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,8,4,power_law_1.01,0.1270246410369873
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,8,4,power_law_1.01,0.13916543960571287
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,8,4,power_law_1.01,0.14142208099365233
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,8,4,power_law_1.01,0.1427494430541992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,8,4,power_law_1.01,0.15151103973388672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,8,4,power_law_1.01,0.16310144424438477
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,8,4,power_law_1.01,0.1650828742980957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,8,4,power_law_1.01,0.1708812713623047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,8,4,power_law_1.01,0.18252288818359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,8,4,power_law_1.01,2.1658816528320313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,8,4,power_law_1.01,0.2052070426940918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,8,4,power_law_1.01,0.24660352706909178
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,8,4,power_law_1.01,0.30740224838256835
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,8,4,power_law_1.01,0.35266048431396485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,8,4,power_law_1.01,0.45028736114501955
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,8,4,power_law_1.01,0.5369753646850586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,8,4,power_law_1.2,0.06397952079772949
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,8,4,power_law_1.01,0.153951997756958
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,8,4,power_law_1.01,2.9058380126953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,8,4,power_law_1.01,0.7664959716796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,8,4,power_law_1.2,0.07635072231292725
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,8,4,power_law_1.2,0.0859712028503418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,8,4,power_law_1.2,0.09908608436584473
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,8,4,power_law_1.01,0.9752909088134766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,8,4,power_law_1.2,0.12608384132385253
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,8,4,power_law_1.2,0.13309696197509766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,8,4,power_law_1.2,0.16893312454223633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,8,4,power_law_1.2,0.2184102439880371
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,8,4,power_law_1.2,0.21913471221923828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,8,4,power_law_1.2,0.22650495529174805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,8,4,power_law_1.2,0.2332262420654297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,8,4,power_law_1.01,1.4447283935546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,8,4,power_law_1.2,0.2354521560668945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,8,4,power_law_1.2,0.23919488906860353
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,8,4,power_law_1.2,0.24356992721557616
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,8,4,power_law_1.2,0.2544051170349121
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,8,4,power_law_1.01,1.8524082946777345
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,8,4,power_law_1.2,0.26961536407470704
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,8,4,power_law_1.2,0.3039782333374023
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,8,4,power_law_1.2,0.3582207870483399
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,8,4,power_law_1.2,0.44685951232910154
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,8,4,power_law_1.2,0.17099775314331053
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,8,4,power_law_1.2,0.5325657653808593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,8,4,power_law_1.2,0.0407539176940918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,8,4,power_law_1.2,0.7254131317138672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,8,4,power_law_1.2,0.8763404846191406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,8,4,power_law_1.2,0.054789118766784675
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,8,4,power_law_1.2,0.06059904098510742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,8,4,power_law_1.2,0.07185664176940917
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,8,4,power_law_1.2,1.2437337493896483
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,8,4,power_law_1.2,0.08690303802490235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,8,4,power_law_1.2,0.0917311954498291
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,8,4,power_law_1.2,0.11168512344360351
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,8,4,power_law_1.2,0.12339327812194825
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,8,4,power_law_1.2,1.5476083374023437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,8,4,power_law_1.2,0.13612159729003906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,8,4,power_law_1.2,0.14204928398132324
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,8,4,power_law_1.2,0.14398079872131347
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,8,4,power_law_1.2,0.15530240058898925
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,8,4,power_law_1.2,0.15585536003112793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,8,4,power_law_1.2,0.16040447235107422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,8,4,power_law_1.2,0.16711936950683595
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,8,4,power_law_1.2,0.17182464599609376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,8,4,power_law_1.2,0.1859379196166992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,8,4,power_law_1.2,0.22105344772338867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,8,4,power_law_1.2,0.2555955123901367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,8,4,power_law_1.2,0.31126272201538085
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,8,4,power_law_1.2,0.35722110748291014
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,8,4,power_law_1.2,0.4640371322631836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,8,4,power_law_1.2,0.5399539184570312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,16,2,balanced,0.038420479297637936
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,8,4,power_law_1.2,3.101643371582031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,16,2,balanced,0.056294398307800295
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,16,2,balanced,0.06508416175842285
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,8,4,power_law_1.2,0.7897920227050781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,16,2,balanced,0.08235648155212402
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,16,2,balanced,0.12174464225769044
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,8,4,power_law_1.2,0.9895629119873046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,16,2,balanced,0.1972787284851074
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,8,4,power_law_1.2,2.3703680419921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,16,2,balanced,0.19780223846435546
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,16,2,balanced,0.20008447647094724
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,16,2,balanced,0.20247936248779297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,16,2,balanced,0.20705408096313477
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,16,2,balanced,0.2547366333007813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,16,2,balanced,0.2587980842590332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,16,2,balanced,0.26461311340332033
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,16,2,balanced,0.26835456848144534
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,8,4,power_law_1.2,1.4912652587890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,16,2,balanced,0.27746688842773437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,16,2,balanced,0.29637376785278324
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,16,2,balanced,0.3167027282714844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,16,2,balanced,0.35926910400390627
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,8,4,power_law_1.2,1.9401011657714844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,16,2,balanced,0.40593406677246097
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,16,2,balanced,0.19521408081054686
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,16,2,balanced,0.656833267211914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,16,2,balanced,0.03185663938522339
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,16,2,balanced,0.252807674407959
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,16,2,balanced,0.7746739196777344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,16,2,balanced,0.0362175989151001
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,16,2,balanced,0.04747392177581787
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,16,2,balanced,0.0653324794769287
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,16,2,balanced,0.09979519844055176
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,16,2,balanced,0.16902784347534178
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,16,2,balanced,0.16971775054931643
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,16,2,balanced,1.1264832305908203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,16,2,balanced,0.171910400390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,16,2,balanced,0.17308416366577148
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,16,2,balanced,0.17457408905029298
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,16,2,balanced,0.1765977668762207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,16,2,balanced,0.22341632843017578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,16,2,balanced,0.22556671142578125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,16,2,balanced,1.4679193115234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,16,2,balanced,0.22778495788574218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,16,2,balanced,0.23469055175781248
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,16,2,balanced,0.23566719055175783
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,16,2,balanced,0.24486656188964845
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,16,2,balanced,0.2617036819458008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,16,2,balanced,0.2789401626586914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,16,2,balanced,0.3178623962402344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,16,2,balanced,0.35666431427001954
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,16,2,balanced,2.0729139709472655
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,16,2,balanced,0.43161598205566404
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,16,2,balanced,0.509030418395996
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,16,2,power_law_1.01,0.05640448093414306
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,16,2,power_law_1.01,0.06411263942718506
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,16,2,balanced,0.7981273651123046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,16,2,power_law_1.01,0.07365375995635987
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,16,2,balanced,2.7195648193359374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,16,2,balanced,0.9587391662597657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,16,2,power_law_1.01,0.14185215950012206
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,16,2,power_law_1.01,0.1659071922302246
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,16,2,power_law_1.01,0.17124351501464843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,16,2,balanced,1.4038156127929686
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,16,2,power_law_1.01,0.1770124816894531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,16,2,power_law_1.01,0.18280832290649413
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,16,2,power_law_1.01,0.227193603515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,16,2,power_law_1.01,0.22605567932128906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,16,2,balanced,1.8586265563964843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,16,2,power_law_1.01,0.09033344268798829
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,16,2,power_law_1.01,0.23469568252563477
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,16,2,power_law_1.01,0.23810815811157227
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,16,2,power_law_1.01,0.11394047737121582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,16,2,power_law_1.01,0.2443187141418457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,16,2,power_law_1.01,0.26067968368530275
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,16,2,power_law_1.01,0.2895142364501953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,16,2,power_law_1.01,0.32391422271728515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,16,2,power_law_1.01,0.16964096069335938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,16,2,power_law_1.01,0.4126476669311524
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,16,2,power_law_1.01,0.5070873641967774
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,16,2,power_law_1.01,0.6807872009277344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,16,2,power_law_1.01,0.03512576103210449
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,16,2,power_law_1.01,0.8229875183105468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,16,2,power_law_1.01,0.04646656036376953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,16,2,power_law_1.01,0.058662400245666504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,16,2,power_law_1.01,0.0717849588394165
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,16,2,power_law_1.01,1.1689433288574218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,16,2,power_law_1.01,0.09441408157348632
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,16,2,power_law_1.01,0.10777855873107911
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,16,2,power_law_1.01,0.13749247550964355
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,16,2,power_law_1.01,0.1456934356689453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,16,2,power_law_1.01,1.4761804199218749
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,16,2,power_law_1.01,0.14648447990417482
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,16,2,power_law_1.01,0.14833791732788087
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,16,2,power_law_1.01,0.1967897605895996
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,16,2,power_law_1.01,0.2105190467834473
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,16,2,power_law_1.01,0.2124736022949219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,16,2,power_law_1.01,0.22190719604492187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,16,2,power_law_1.01,2.1895411682128905
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,16,2,power_law_1.01,0.24947584152221677
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,16,2,power_law_1.01,0.2831513595581055
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,16,2,power_law_1.01,0.3584140777587891
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,16,2,power_law_1.01,0.40575614929199216
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,16,2,power_law_1.01,2.7382104492187502
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,16,2,power_law_1.01,0.5116070556640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,16,2,power_law_1.01,0.19593727111816406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,16,2,power_law_1.01,0.20248447418212892
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,16,2,power_law_1.01,0.6142335891723633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,16,2,power_law_1.2,0.055313920974731444
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,16,2,power_law_1.2,0.06440832138061522
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,16,2,power_law_1.2,0.07016831874847412
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,16,2,power_law_1.2,0.08739583969116212
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,16,2,power_law_1.01,0.855203857421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,16,2,power_law_1.2,0.11185407638549805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,16,2,power_law_1.01,0.14525055885314941
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,16,2,power_law_1.2,0.1316160011291504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,16,2,power_law_1.2,0.16368640899658202
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,16,2,power_law_1.01,1.0704051208496093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,16,2,power_law_1.2,0.16551040649414062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,16,2,power_law_1.2,0.17176448822021484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,16,2,power_law_1.2,0.1765337562561035
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,16,2,power_law_1.2,0.18495359420776367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,16,2,power_law_1.2,0.21993087768554687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,16,2,power_law_1.2,0.23256576538085935
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,16,2,power_law_1.2,0.24296960830688477
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,16,2,power_law_1.01,1.5574246215820313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,16,2,power_law_1.2,0.25084159851074217
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,16,2,power_law_1.2,0.26674303054809567
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,16,2,power_law_1.2,0.2941644859313965
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,16,2,power_law_1.2,0.3302041625976563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,16,2,power_law_1.01,1.9729356384277346
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,16,2,power_law_1.2,0.4409535980224609
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,16,2,power_law_1.2,0.5221222305297852
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,16,2,power_law_1.2,0.684985580444336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,16,2,power_law_1.2,0.03486848115921021
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,16,2,power_law_1.2,0.8403404998779298
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,16,2,power_law_1.2,0.04585728168487549
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,16,2,power_law_1.2,0.05367936134338379
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,16,2,power_law_1.2,0.07046144008636475
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,16,2,power_law_1.2,1.2148941040039063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,16,2,power_law_1.2,0.08883456230163575
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,16,2,power_law_1.2,0.1331276798248291
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,16,2,power_law_1.2,1.5079347229003905
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,16,2,power_law_1.2,0.14404224395751952
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,16,2,power_law_1.2,0.15073920249938966
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,16,2,power_law_1.2,0.15299839973449708
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,16,2,power_law_1.2,0.22758272171020505
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,16,2,power_law_1.2,0.19361024856567383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,16,2,power_law_1.2,0.19602304458618164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,16,2,power_law_1.2,0.20033536911010744
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,16,2,power_law_1.2,0.20859136581420898
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,16,2,power_law_1.2,2.2255871582031252
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,16,2,power_law_1.2,0.208590087890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,16,2,power_law_1.2,0.23590400695800778
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,16,2,power_law_1.2,0.10209407806396484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,16,2,power_law_1.2,0.25195520401000976
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,16,2,power_law_1.2,0.14505472183227539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,16,2,power_law_1.2,0.3122879981994629
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,16,2,power_law_1.2,0.3699148941040039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,16,2,power_law_1.2,2.8191015624999998
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,16,2,power_law_1.2,0.4187916946411133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,32,1,balanced,0.050391039848327636
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,16,2,power_law_1.2,0.5258611297607422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,16,2,power_law_1.2,0.628408317565918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,32,1,balanced,0.06732351779937744
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,32,1,balanced,0.07786623954772949
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,32,1,balanced,0.09936832427978516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,32,1,balanced,0.14506175994873047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,32,1,balanced,0.2479020881652832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,16,2,power_law_1.2,1.0855948638916015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,32,1,balanced,0.2482137680053711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,32,1,balanced,0.24937152862548828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,32,1,balanced,0.2527411270141601
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,32,1,balanced,0.2540204811096191
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,32,1,balanced,0.2587116813659668
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,32,1,balanced,0.2645120048522949
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,16,2,power_law_1.2,1.585218505859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,32,1,balanced,0.27425920486450195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,32,1,balanced,0.28877824783325196
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,32,1,balanced,0.29576831817626953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,32,1,balanced,0.2988947105407715
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,16,2,power_law_1.2,2.0145677185058597
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,32,1,balanced,0.3066348838806152
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,32,1,balanced,0.3251808166503906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,32,1,balanced,0.3428192138671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,32,1,balanced,0.3899590301513672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,32,1,balanced,0.44581825256347657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,32,1,balanced,0.04227968215942383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,32,1,balanced,0.7168083190917969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,32,1,balanced,0.7752690887451171
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,32,1,balanced,0.051376638412475584
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,32,1,balanced,0.06607808113098144
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,32,1,balanced,0.09541631698608398
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,32,1,balanced,0.14953920364379883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,32,1,balanced,1.0745101165771485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,32,1,balanced,0.2672064018249512
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,32,1,balanced,0.2674687957763672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,32,1,balanced,0.2679334449768066
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,32,1,balanced,0.27121088027954104
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,32,1,balanced,1.4510406494140624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,32,1,balanced,0.27245887756347653
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,32,1,balanced,0.27545536041259766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,16,2,power_law_1.2,0.8728665924072265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,32,1,balanced,0.277007999420166
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,32,1,balanced,0.28079872131347655
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,32,1,balanced,0.28593791961669923
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,32,1,balanced,0.37761280059814456
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,32,1,balanced,0.3785971069335937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,32,1,balanced,0.3852870559692383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,32,1,balanced,0.39834945678710937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,32,1,balanced,2.1730828857421876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,32,1,balanced,0.41608833312988286
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,32,1,balanced,0.44909503936767575
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,32,1,balanced,0.4850809478759766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,32,1,balanced,0.5585273742675781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,32,1,balanced,0.6343795013427734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,32,1,power_law_1.01,0.0408512020111084
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,32,1,balanced,2.8549041748046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,32,1,power_law_1.01,0.06692416191101073
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,32,1,power_law_1.01,0.07756095886230469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,32,1,balanced,1.0487142181396485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,32,1,power_law_1.01,0.09947392463684082
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,32,1,power_law_1.01,0.12693440437316894
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,32,1,power_law_1.01,0.1677587127685547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,32,1,balanced,1.2032780456542969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,32,1,power_law_1.01,0.20474304199218749
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,32,1,power_law_1.01,0.21135488510131836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,32,1,power_law_1.01,0.216180477142334
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,32,1,power_law_1.01,0.21929664611816407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,32,1,power_law_1.01,0.2354015922546387
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,32,1,power_law_1.01,0.24553855895996093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,32,1,power_law_1.01,0.2596108818054199
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,32,1,balanced,1.7841357421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,32,1,power_law_1.01,0.28346752166748046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,32,1,power_law_1.01,0.27347904205322265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,32,1,power_law_1.01,0.2776563262939453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,32,1,power_law_1.01,0.2939936065673828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,32,1,power_law_1.01,0.32204734802246093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,32,1,power_law_1.01,0.351607666015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,32,1,power_law_1.01,0.41823871612548824
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,32,1,power_law_1.01,0.5006521606445313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,32,1,balanced,2.375990447998047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,32,1,power_law_1.01,0.7005010986328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,32,1,power_law_1.01,0.042236800193786624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,32,1,power_law_1.01,0.04904640197753906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,32,1,power_law_1.01,0.8297229003906249
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,32,1,power_law_1.01,0.06627903938293457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,32,1,power_law_1.01,0.09460160255432129
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,32,1,power_law_1.01,1.1660467529296876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,32,1,power_law_1.01,0.12539584159851075
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,32,1,power_law_1.01,0.1593734359741211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,32,1,power_law_1.01,0.20801151275634763
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,32,1,power_law_1.01,0.2160576057434082
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,32,1,power_law_1.01,1.5144114685058594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,32,1,power_law_1.01,0.22152704238891602
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,32,1,power_law_1.01,0.2241043281555176
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,32,1,power_law_1.01,0.23527679443359376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,32,1,power_law_1.01,0.2432569694519043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,32,1,power_law_1.01,0.2481817626953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,32,1,power_law_1.01,0.2535750389099121
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,32,1,power_law_1.01,0.34499969482421877
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,32,1,power_law_1.01,0.34689537048339847
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,32,1,power_law_1.01,0.3623993682861328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,32,1,power_law_1.01,2.202645721435547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,32,1,power_law_1.01,0.3847750473022461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,32,1,power_law_1.01,0.42623550415039063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,32,1,power_law_1.01,0.5136748886108398
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,32,1,power_law_1.01,0.5704281616210938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,32,1,power_law_1.01,2.8734451293945313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,32,1,power_law_1.01,0.8582803344726562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,32,1,power_law_1.2,0.040979199409484864
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,32,1,power_law_1.01,1.129778594970703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,32,1,power_law_1.2,0.06697343826293946
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,32,1,power_law_1.2,0.07795263767242432
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,32,1,power_law_1.2,0.09585984230041504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,32,1,power_law_1.01,0.6957376098632813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,32,1,power_law_1.01,1.4317523193359374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,32,1,power_law_1.2,0.12150655746459962
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,32,1,power_law_1.2,0.1599007987976074
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,32,1,power_law_1.2,0.20367551803588868
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,32,1,power_law_1.2,0.21002624511718748
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,32,1,power_law_1.2,0.2155187225341797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,32,1,power_law_1.2,0.21989824295043947
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,32,1,power_law_1.2,0.23401407241821287
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,32,1,power_law_1.2,0.2451852798461914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,32,1,power_law_1.2,0.26144895553588865
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,32,1,power_law_1.2,0.2892038345336914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,32,1,power_law_1.01,2.003396453857422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,32,1,power_law_1.2,0.27940160751342774
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,32,1,power_law_1.2,0.28496192932128905
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,32,1,power_law_1.2,0.2976268768310547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,32,1,power_law_1.2,0.3330950546264648
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,32,1,power_law_1.2,0.36083393096923827
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,32,1,power_law_1.2,0.4301043319702148
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,32,1,power_law_1.01,2.569274291992188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,32,1,power_law_1.2,0.5178195190429687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,32,1,power_law_1.2,0.7202655792236328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,32,1,power_law_1.2,0.04220863819122315
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,32,1,power_law_1.2,0.8475635528564454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,32,1,power_law_1.2,0.06682176113128661
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,32,1,power_law_1.2,0.08996224403381348
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,32,1,power_law_1.2,1.1806336212158204
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,32,1,power_law_1.2,0.1215238380432129
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,32,1,power_law_1.2,0.155284481048584
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,32,1,power_law_1.2,0.2037241554260254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,32,1,power_law_1.2,1.540417938232422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,32,1,power_law_1.2,0.21228864669799802
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,32,1,power_law_1.2,0.21854272842407227
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,32,1,power_law_1.2,0.04898880004882812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,32,1,power_law_1.2,0.22662271499633788
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,32,1,power_law_1.2,0.2329190444946289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,32,1,power_law_1.2,0.24081600189208982
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,32,1,power_law_1.2,0.24865343093872072
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,32,1,power_law_1.2,0.25767871856689456
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,32,1,power_law_1.2,0.3497190475463867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,32,1,power_law_1.2,2.2243238830566403
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,32,1,power_law_1.2,0.35823745727539064
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,32,1,power_law_1.2,0.3668326568603516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,32,1,power_law_1.2,0.4120147323608398
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,32,1,power_law_1.2,0.4659392166137695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,32,1,power_law_1.2,0.5281369781494141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,32,1,power_law_1.2,0.5904684829711914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,32,1,power_law_1.2,2.9035653686523437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,32,1,power_law_1.2,0.7266655731201171
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,32,1,power_law_1.2,0.8842336273193359
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,32,1,power_law_1.2,1.1481292724609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,1,32,balanced,0.0898624038696289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,32,1,power_law_1.2,1.4602706909179688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,1,32,balanced,0.16399744033813476
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,1,32,balanced,0.16610687255859374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,32,1,power_law_1.2,2.0186810302734375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,1,32,balanced,0.16165248870849608
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,1,32,balanced,0.1744972801208496
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,32,1,power_law_1.2,2.595364990234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,1,32,balanced,0.19751552581787107
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,1,32,balanced,0.17451391220092774
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,1,32,balanced,0.17852415084838866
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,1,32,balanced,0.1820416069030762
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,1,32,balanced,0.06138239860534668
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,1,32,balanced,0.18644351959228517
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,1,32,balanced,0.20834175109863282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,1,32,balanced,0.046210560798645016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,1,32,balanced,0.22537471771240236
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,1,32,balanced,0.23206783294677735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,1,32,balanced,0.04644608020782471
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,1,32,balanced,0.046369280815124515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,1,32,balanced,0.16134912490844727
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,1,32,balanced,0.2911078453063965
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,1,32,balanced,0.3886195373535156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,1,32,balanced,0.17045631408691406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,1,32,balanced,0.03639935970306397
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,1,32,balanced,0.6265075302124024
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,1,32,balanced,0.036796159744262695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,1,32,balanced,0.7997478485107422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,1,32,balanced,0.26013311386108395
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,1,32,balanced,0.059278078079223626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,1,32,balanced,0.09563648223876953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,1,32,balanced,0.09855744361877442
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,1,32,balanced,0.0971827220916748
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,1,32,balanced,0.100633602142334
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,1,32,balanced,0.09974143981933595
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,1,32,balanced,0.10239744186401367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,1,32,balanced,0.104335355758667
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,1,32,balanced,0.17171455383300782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,1,32,balanced,0.10980992317199707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,1,32,balanced,0.1144115161895752
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,1,32,balanced,0.11554944038391113
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,1,32,balanced,0.12022527694702148
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,1,32,balanced,0.12797311782836912
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,1,32,balanced,0.17693439483642578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,1,32,balanced,0.143056640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,1,32,balanced,0.16161663055419923
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,1,32,balanced,0.10129664421081544
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,1,32,balanced,0.19260671615600586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,1,32,balanced,0.22423295974731444
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,1,32,balanced,0.3027136039733887
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,1,32,balanced,0.4934592056274414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,1,32,balanced,0.6388582229614258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,1,32,power_law_1.01,0.1648921585083008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,1,32,power_law_1.01,0.15637760162353515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,1,32,power_law_1.01,0.1631590461730957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,1,32,balanced,0.11057791709899903
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,1,32,power_law_1.01,0.14968959808349608
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,1,32,power_law_1.01,0.1650739288330078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,1,32,balanced,0.03689984083175659
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,1,32,power_law_1.01,0.1694924736022949
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,1,32,power_law_1.01,0.15922047615051269
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,1,32,power_law_1.01,0.1557094383239746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,1,32,power_law_1.01,0.12660736083984375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,1,32,power_law_1.01,0.16128639221191404
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,1,32,power_law_1.01,0.16007423400878906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,1,32,balanced,0.45517055511474613
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,1,32,power_law_1.01,0.18628992080688475
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,1,32,power_law_1.01,0.19274751663208006
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,1,32,power_law_1.01,0.2665100860595703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,1,32,power_law_1.01,0.2828505516052246
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,1,32,power_law_1.01,0.30667007446289063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,1,32,power_law_1.01,0.3984947204589844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,1,32,power_law_1.01,0.4752345657348632
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,1,32,balanced,0.047918081283569336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,1,32,power_law_1.01,0.8463692474365235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,1,32,power_law_1.01,0.148723201751709
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,1,32,power_law_1.01,1.1629055786132814
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,1,32,balanced,0.3530316925048828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,1,32,power_law_1.01,0.10120832443237304
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,1,32,power_law_1.01,1.6343014526367188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,1,32,power_law_1.01,0.09727744102478028
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,1,32,power_law_1.01,1.762809600830078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,1,32,power_law_1.01,0.08632960319519042
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,1,32,power_law_1.01,0.1475494384765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,1,32,power_law_1.01,0.09206527709960938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,1,32,power_law_1.01,0.09106304168701171
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,1,32,power_law_1.01,0.09752703666687011
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,1,32,power_law_1.01,0.16862848281860351
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,1,32,power_law_1.01,0.10106368064880371
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,1,32,power_law_1.01,0.10553088188171386
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,1,32,power_law_1.01,0.09546112060546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,1,32,power_law_1.01,0.07688704013824463
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,1,32,power_law_1.01,0.5530559921264648
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,1,32,power_law_1.01,0.11249664306640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,1,32,power_law_1.01,0.1260646438598633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,1,32,power_law_1.01,0.16919551849365236
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,1,32,power_law_1.01,0.17741056442260744
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,1,32,power_law_1.01,0.09936639785766602
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,1,32,power_law_1.01,0.20769023895263672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,1,32,power_law_1.01,0.10857600212097167
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,1,32,power_law_1.01,0.28198144912719725
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,1,32,power_law_1.01,0.16228864669799806
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,1,32,power_law_1.01,0.6394534301757813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,1,32,power_law_1.01,0.8781887817382813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,1,32,power_law_1.2,0.16095104217529296
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,1,32,power_law_1.01,1.565727996826172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,1,32,power_law_1.01,1.2098099517822267
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,1,32,power_law_1.01,0.09683199882507323
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,1,32,power_law_1.2,0.15802751541137697
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,1,32,power_law_1.2,0.11811455726623535
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,1,32,power_law_1.01,0.09362688064575195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,1,32,power_law_1.2,0.16510080337524413
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,1,32,power_law_1.2,0.13739775657653808
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,1,32,power_law_1.2,0.142871036529541
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,1,32,power_law_1.2,0.160894718170166
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,1,32,power_law_1.2,0.15406335830688475
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,1,32,power_law_1.2,0.15748736381530762
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,1,32,power_law_1.2,0.17472511291503906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,1,32,power_law_1.2,0.1935270309448242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,1,32,power_law_1.2,0.1157094383239746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,1,32,power_law_1.2,0.1929804801940918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,1,32,power_law_1.2,0.2423334312438965
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,1,32,power_law_1.2,0.2689753532409668
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,1,32,power_law_1.2,0.2940415954589844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,1,32,power_law_1.01,0.10812671661376953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,1,32,power_law_1.2,0.33192703247070315
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,1,32,power_law_1.2,0.4315801620483398
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,1,32,power_law_1.2,0.5676198577880859
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,1,32,power_law_1.2,0.7930393218994141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,1,32,power_law_1.2,1.0910655975341796
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,1,32,power_law_1.2,1.4066099548339843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,1,32,power_law_1.2,2.266288604736328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,1,32,power_law_1.2,0.103056640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,1,32,power_law_1.2,0.09722623825073243
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,1,32,power_law_1.2,0.09509759902954101
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,1,32,power_law_1.2,0.1295193576812744
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,1,32,power_law_1.2,0.09406208038330079
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,1,32,power_law_1.2,0.0832857608795166
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,1,32,power_law_1.2,0.09022591590881349
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,1,32,power_law_1.2,0.0857817554473877
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,1,32,power_law_1.2,0.0971622371673584
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,1,32,power_law_1.2,0.09694463729858398
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,1,32,power_law_1.2,0.10433919906616211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,1,32,power_law_1.2,0.1102937602996826
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,1,32,power_law_1.2,3.254429321289062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,1,32,power_law_1.2,0.10796799659729003
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,1,32,power_law_1.2,0.11746944427490233
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,1,32,power_law_1.2,0.14684415817260743
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,1,32,power_law_1.2,0.1578598403930664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,1,32,power_law_1.2,0.17295616149902343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,1,32,power_law_1.2,0.19242752075195313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,1,32,power_law_1.2,0.2402521514892578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,1,32,power_law_1.2,0.12101632118225099
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,1,32,power_law_1.2,0.08132096290588378
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,1,32,power_law_1.2,0.1611724853515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,1,32,power_law_1.2,0.8615551757812501
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,1,32,power_law_1.2,1.024111328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,1,32,power_law_1.01,0.31986303329467775
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,1,32,power_law_1.2,1.632048645019531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,1,32,power_law_1.01,0.45804672241210936
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,2,16,balanced,0.03854079961776734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,2,16,balanced,0.057731838226318354
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,2,16,balanced,0.09299200057983399
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,1,32,power_law_1.2,2.017242889404297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,1,32,power_law_1.2,0.10101504325866699
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,2,16,balanced,0.1670412826538086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,2,16,balanced,0.1681996726989746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,2,16,balanced,0.16943359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,2,16,balanced,0.19874176025390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,2,16,balanced,0.17617151260375977
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,2,16,balanced,0.18769023895263673
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,2,16,balanced,0.17408128738403322
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,2,16,balanced,0.18282751083374021
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,2,16,balanced,0.18485631942749023
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,2,16,balanced,0.02888959884643555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,2,16,balanced,0.029250559806823728
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,2,16,balanced,0.21119232177734376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,2,16,balanced,0.23227008819580078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,2,16,balanced,0.1721625518798828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,2,16,balanced,0.2977203178405762
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,2,16,balanced,0.16746368408203124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,2,16,balanced,0.35819393157958984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,2,16,balanced,0.4826496124267578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,2,16,balanced,0.5755532836914062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,2,16,balanced,0.8310502624511719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,2,16,balanced,0.030624001026153563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,2,16,balanced,1.0943475341796876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,2,16,balanced,0.030533120632171628
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,2,16,balanced,0.03464447975158692
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,2,16,balanced,0.04507264137268067
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,2,16,balanced,0.061954560279846194
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,2,16,balanced,0.09982080459594726
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,2,16,balanced,0.11034751892089845
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,2,16,balanced,0.11149439811706544
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,2,16,balanced,0.11332863807678221
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,2,16,balanced,0.11306752204895018
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,2,16,balanced,0.11646207809448242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,2,16,balanced,0.11879424095153808
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,2,16,balanced,0.1858086395263672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,2,16,balanced,0.1225011157989502
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,2,16,balanced,0.11593728065490723
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,2,16,balanced,0.20158079147338864
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,2,16,balanced,0.11843968391418458
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,2,16,balanced,0.1218496036529541
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,2,16,balanced,0.1273459243774414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,2,16,balanced,0.13919103622436524
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,2,16,balanced,0.1535103988647461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,2,16,balanced,0.17874687194824218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,2,16,balanced,0.19884031295776367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,2,16,balanced,0.24290815353393555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,2,16,balanced,0.2882854461669922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,2,16,balanced,0.25273216247558594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,2,16,balanced,0.4065164947509766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,2,16,power_law_1.01,0.08851455688476563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,2,16,balanced,0.496099853515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,2,16,balanced,0.7084748840332031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,2,16,power_law_1.01,0.1624844741821289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,2,16,balanced,0.9240345764160157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,2,16,power_law_1.01,0.1564083194732666
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,1,32,power_law_1.2,0.29267967224121094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,2,16,power_law_1.01,0.13870847702026368
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,2,16,power_law_1.01,0.1508070373535156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,1,32,power_law_1.2,0.41883262634277346
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,2,16,power_law_1.01,0.15824383735656739
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,2,16,power_law_1.01,0.17570047378540038
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,2,16,power_law_1.01,0.1886899185180664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,1,32,power_law_1.2,0.5029926300048828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,2,16,power_law_1.01,0.15837568283081055
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,2,16,power_law_1.01,0.1313203239440918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,2,16,power_law_1.01,0.1435366439819336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,2,16,power_law_1.01,0.22305152893066404
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,2,16,power_law_1.01,0.1535961627960205
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,2,16,power_law_1.01,0.3089331245422363
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,2,16,power_law_1.01,0.1661644744873047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,2,16,power_law_1.01,0.4363289642333984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,2,16,power_law_1.01,0.7453670501708984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,2,16,power_law_1.01,0.20551807403564454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,2,16,power_law_1.01,0.9881075286865235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,2,16,power_law_1.01,0.20688640594482424
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,2,16,power_law_1.01,0.060186882019042966
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,2,16,power_law_1.01,0.09565823554992677
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,2,16,power_law_1.01,1.4467800903320314
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,2,16,power_law_1.01,0.09605376243591308
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,2,16,power_law_1.01,0.28781312942504883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,2,16,power_law_1.01,0.10211456298828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,2,16,power_law_1.01,0.09298432350158692
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,2,16,power_law_1.01,0.08822784423828126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,2,16,power_law_1.01,0.09283455848693847
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,2,16,power_law_1.01,0.09813759803771974
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,2,16,power_law_1.01,0.0983679962158203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,2,16,power_law_1.01,1.9150746154785157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,2,16,power_law_1.01,0.10051584243774414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,2,16,power_law_1.01,0.10783488273620605
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,2,16,power_law_1.01,0.11220352172851562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,2,16,power_law_1.01,0.1088806438446045
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,2,16,power_law_1.01,0.11850367546081544
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,2,16,power_law_1.01,0.11552895545959471
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,2,16,power_law_1.01,0.11766400337219238
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,2,16,power_law_1.01,0.11859199523925781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,2,16,power_law_1.01,0.1645132827758789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,2,16,power_law_1.01,0.1898636817932129
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,2,16,power_law_1.01,0.20959360122680665
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,2,16,power_law_1.01,0.34217601776123047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,2,16,power_law_1.01,0.2436467170715332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,2,16,power_law_1.01,0.3056409645080566
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,2,16,power_law_1.01,0.4452851104736328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,2,16,power_law_1.01,0.5893670272827148
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,2,16,power_law_1.2,0.08819199562072753
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,2,16,power_law_1.01,0.7747622680664062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,2,16,power_law_1.2,0.15375743865966798
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,2,16,power_law_1.01,1.1064422607421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,2,16,power_law_1.2,0.11860608100891112
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,2,16,power_law_1.01,0.1948646354675293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,2,16,power_law_1.01,1.40932861328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,2,16,power_law_1.2,0.15901056289672852
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,2,16,power_law_1.2,0.1568832015991211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,2,16,power_law_1.2,0.17817855834960938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,2,16,power_law_1.01,0.20497663497924803
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,2,16,power_law_1.2,0.19238527297973634
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,2,16,power_law_1.2,0.19117439270019532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,2,16,power_law_1.2,0.1983513641357422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,2,16,power_law_1.2,0.20728576660156248
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,2,16,power_law_1.2,0.21811328887939455
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,2,16,power_law_1.2,0.2590080070495605
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,2,16,power_law_1.2,0.2945523262023926
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,2,16,power_law_1.01,0.5733478546142579
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,2,16,power_law_1.2,0.3719001770019531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,2,16,power_law_1.2,0.14314240455627442
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,2,16,power_law_1.2,0.5011110305786133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,2,16,power_law_1.2,0.16159616470336916
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,2,16,power_law_1.2,0.16526975631713867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,2,16,power_law_1.2,0.6621234893798829
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,2,16,power_law_1.2,0.1500825595855713
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,2,16,power_law_1.2,0.06011648178100586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,2,16,power_law_1.2,0.09388928413391114
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,2,16,power_law_1.2,1.1003097534179687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,2,16,power_law_1.2,0.08509951591491699
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,2,16,power_law_1.2,0.09709952354431153
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,2,16,power_law_1.2,0.0979097557067871
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,2,16,power_law_1.2,0.1935526466369629
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,2,16,power_law_1.2,1.7850880432128906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,2,16,power_law_1.2,0.08056447982788087
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,2,16,power_law_1.2,0.09180031776428223
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,2,16,power_law_1.2,0.09337599754333496
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,2,16,power_law_1.2,0.09972224235534669
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,2,16,power_law_1.2,0.09901951789855958
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,2,16,power_law_1.2,0.10950655937194824
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,2,16,power_law_1.2,0.11040384292602538
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,2,16,power_law_1.2,0.10743680000305175
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,2,16,power_law_1.2,0.11400832176208495
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,2,16,power_law_1.2,0.12118399620056153
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,2,16,power_law_1.2,2.6922113037109376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,2,16,power_law_1.2,0.13240063667297364
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,2,16,power_law_1.2,0.17230335235595703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,2,16,power_law_1.2,0.18900863647460936
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,2,16,power_law_1.2,0.22221567153930666
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,2,16,power_law_1.2,0.2587788772583008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,2,16,power_law_1.2,0.4005606460571289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,2,16,power_law_1.2,0.11410047531127929
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,2,16,power_law_1.2,0.4399820709228516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,4,8,balanced,0.02964992046356201
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,2,16,power_law_1.2,0.14285823822021484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,2,16,power_law_1.2,0.6712844848632813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,4,8,balanced,0.04767231941223145
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,4,8,balanced,0.05760511875152587
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,4,8,balanced,0.06904831886291504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,4,8,balanced,0.09828479766845703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,4,8,balanced,0.16296447753906249
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,4,8,balanced,0.21738111495971682
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,4,8,balanced,0.2189580726623535
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,4,8,balanced,0.22220800399780272
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,2,16,power_law_1.2,0.31731840133666994
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,4,8,balanced,0.225980167388916
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,2,16,power_law_1.2,1.3728012084960937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,4,8,balanced,0.21597055435180662
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,4,8,balanced,0.22036224365234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,4,8,balanced,0.22395519256591795
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,4,8,balanced,0.23510015487670896
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,4,8,balanced,0.22940416336059571
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,4,8,balanced,0.2420134353637695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,4,8,balanced,0.25004671096801756
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,2,16,power_law_1.2,0.8482418823242188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,2,16,power_law_1.2,1.7705369567871094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,4,8,balanced,0.2723891258239746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,4,8,balanced,0.3167500877380371
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,4,8,balanced,0.35743232727050783
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,4,8,balanced,0.40758399963378905
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,4,8,balanced,0.579947509765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,4,8,balanced,0.6832166290283203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,4,8,balanced,0.03002239942550659
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,4,8,balanced,0.030804479122161867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,4,8,balanced,0.9371212768554689
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,4,8,balanced,0.03667327880859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,4,8,balanced,0.04937983989715576
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,4,8,balanced,0.06560512065887451
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,4,8,balanced,1.2104370880126953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,4,8,balanced,0.10665344238281249
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,4,8,balanced,0.12182144165039062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,4,8,balanced,0.12415231704711913
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,4,8,balanced,0.126812162399292
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,4,8,balanced,0.1298316764831543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,4,8,balanced,0.12128255844116212
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,4,8,balanced,0.12365440368652343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,4,8,balanced,0.12572159767150878
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,4,8,balanced,0.13118207931518555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,4,8,balanced,1.7393370056152342
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,4,8,balanced,0.1360537624359131
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,4,8,balanced,0.14242688179016114
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,4,8,balanced,0.1543295955657959
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,4,8,balanced,0.17457151412963867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,4,8,balanced,0.194202880859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,4,8,balanced,0.2349504089355469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,4,8,balanced,0.27379072189331055
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,4,8,balanced,0.3531468963623047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,4,8,balanced,2.263019561767578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,4,8,balanced,0.43322494506835935
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,4,8,power_law_1.01,0.06768383979797363
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,4,8,power_law_1.01,0.0946291160583496
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,4,8,balanced,0.6308697509765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,4,8,power_law_1.01,0.11566207885742188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,4,8,power_law_1.01,0.12354432106018068
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,4,8,power_law_1.01,0.14069503784179688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,4,8,power_law_1.01,0.12031488418579102
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,4,8,power_law_1.01,0.18820735931396484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,4,8,power_law_1.01,0.18871295928955079
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,4,8,balanced,1.1464793395996093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,4,8,power_law_1.01,0.19543296813964844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,4,8,power_law_1.01,0.19772415161132811
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,4,8,power_law_1.01,0.20363904953002931
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,4,8,power_law_1.01,0.2079859161376953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,4,8,balanced,1.5039347839355468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,4,8,power_law_1.01,0.20155775070190432
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,4,8,power_law_1.01,0.20832511901855472
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,4,8,power_law_1.01,0.21956735610961914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,4,8,power_law_1.01,0.22778112411499024
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,4,8,power_law_1.01,0.24079872131347657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,4,8,balanced,0.7862400054931641
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,4,8,power_law_1.01,0.3095142364501953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,4,8,power_law_1.01,0.27858943939208985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,4,8,power_law_1.01,0.4109107208251953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,4,8,power_law_1.01,0.5133683013916015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,4,8,power_law_1.01,0.6855564880371093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,4,8,power_law_1.01,0.04763519763946533
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,4,8,power_law_1.01,0.8096972656249999
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,4,8,power_law_1.01,0.06296832084655761
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,4,8,power_law_1.01,0.07741951942443848
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,4,8,power_law_1.01,0.08263551712036132
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,4,8,power_law_1.01,1.0722124481201172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,4,8,power_law_1.01,0.08955264091491699
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,4,8,power_law_1.01,0.08054911613464355
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,4,8,power_law_1.01,1.399471435546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,4,8,power_law_1.01,0.10317567825317382
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,4,8,power_law_1.01,0.10934656143188477
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,4,8,power_law_1.01,0.10533503532409667
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,4,8,power_law_1.01,0.11603967666625976
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,4,8,power_law_1.01,0.11314559936523438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,4,8,power_law_1.01,0.11571200370788574
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,4,8,power_law_1.01,2.0254425048828124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,4,8,power_law_1.01,0.12362112045288084
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,2,16,power_law_1.2,1.0048716735839842
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,4,8,power_law_1.01,0.13167360305786133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,4,8,power_law_1.01,0.13202943801879882
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,4,8,power_law_1.01,0.16042623519897461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,4,8,power_law_1.01,0.20373888015747071
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,4,8,power_law_1.01,0.24470144271850586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,4,8,power_law_1.01,0.0964083194732666
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,4,8,power_law_1.01,2.564747619628906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,4,8,power_law_1.01,0.2762303924560547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,4,8,power_law_1.01,0.3332992172241211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,4,8,power_law_1.01,0.11355648040771485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,4,8,power_law_1.2,0.06803967952728271
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,4,8,power_law_1.01,0.4361446380615234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,4,8,power_law_1.2,0.09330559730529785
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,4,8,power_law_1.2,0.09423616409301758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,4,8,power_law_1.2,0.11488256454467774
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,4,8,power_law_1.2,0.1425920009613037
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,4,8,power_law_1.01,0.7965760040283203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,4,8,power_law_1.2,0.13882880210876464
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,4,8,power_law_1.2,0.18460031509399416
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,4,8,power_law_1.01,1.1160934448242188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,4,8,power_law_1.2,0.18872703552246092
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,4,8,power_law_1.2,0.18823680877685547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,4,8,power_law_1.2,0.19770368576049804
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,4,8,power_law_1.2,0.20327295303344725
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,4,8,power_law_1.01,1.4070477294921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,4,8,power_law_1.2,0.20060543060302732
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,4,8,power_law_1.2,0.20342016220092773
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,4,8,power_law_1.2,0.2173516845703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,4,8,power_law_1.01,0.6362035369873047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,4,8,power_law_1.2,0.2298918342590332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,4,8,power_law_1.2,0.25190528869628903
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,4,8,power_law_1.2,0.31116287231445316
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,4,8,power_law_1.2,0.33753089904785155
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,4,8,power_law_1.2,0.40125312805175783
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,4,8,power_law_1.2,0.5248947143554688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,4,8,power_law_1.2,0.6919462585449219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,4,8,power_law_1.2,0.21483392715454103
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,4,8,power_law_1.2,0.823683853149414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,4,8,power_law_1.2,0.04761727809906006
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,4,8,power_law_1.2,0.061585922241210934
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,4,8,power_law_1.2,0.06381824016571044
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,4,8,power_law_1.2,1.1794764709472656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,4,8,power_law_1.2,0.07726336002349854
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,4,8,power_law_1.2,0.08444160461425781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,4,8,power_law_1.2,1.4270425415039063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,4,8,power_law_1.2,0.06891776084899902
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,4,8,power_law_1.2,0.09741439819335937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,4,8,power_law_1.2,0.10343168258666993
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,4,8,power_law_1.2,0.10321408271789552
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,4,8,power_law_1.2,0.10743424415588379
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,4,8,power_law_1.2,0.11466239929199218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,4,8,power_law_1.2,0.11652864456176756
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,4,8,power_law_1.2,0.12085887908935547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,4,8,power_law_1.2,0.12252032279968261
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,4,8,power_law_1.2,0.13145600318908693
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,4,8,power_law_1.2,0.14957823753356933
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,4,8,power_law_1.2,0.17843839645385742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,4,8,power_law_1.2,0.204899845123291
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,4,8,power_law_1.2,2.222337341308594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,4,8,power_law_1.2,0.2511961555480957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,4,8,power_law_1.2,0.29092607498168943
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,4,8,power_law_1.2,0.38001022338867185
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,4,8,power_law_1.2,0.45451007843017577
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,8,4,balanced,0.028860158920288086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,8,4,balanced,0.04811264038085937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,4,8,power_law_1.2,0.6307660675048827
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,8,4,balanced,0.056939520835876466
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,8,4,balanced,0.07107071876525879
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,4,8,power_law_1.2,0.823262710571289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,8,4,balanced,0.09762175559997559
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,8,4,balanced,0.15953791618347168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,8,4,balanced,0.16181631088256837
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,4,8,power_law_1.2,3.1128436279296876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,8,4,balanced,0.16412416458129883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,8,4,balanced,0.20878463745117187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,8,4,balanced,0.20987007141113284
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,4,8,power_law_1.2,1.2104358673095703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,8,4,balanced,0.2122700881958008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,8,4,balanced,0.21685247421264647
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,8,4,balanced,0.21837696075439453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,8,4,balanced,0.22284927368164062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,8,4,balanced,0.22659328460693356
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,8,4,balanced,0.24076160430908203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,8,4,balanced,0.23172992706298828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,8,4,balanced,0.26065792083740236
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,4,8,power_law_1.2,0.11277695655822755
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,8,4,balanced,0.2810073661804199
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,4,8,power_law_1.2,1.5934515380859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,8,4,balanced,0.3303744125366211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,8,4,balanced,0.37713153839111324
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,8,4,balanced,0.028744959831237794
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,8,4,balanced,0.03032320022583008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,8,4,balanced,0.5692185592651368
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,8,4,balanced,0.03904383897781372
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,8,4,balanced,0.6918118286132813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,8,4,balanced,0.05149439811706543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,8,4,balanced,0.07088768005371093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,8,4,balanced,0.11726719856262206
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,8,4,balanced,0.1161676788330078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,8,4,balanced,0.11951359748840332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,8,4,balanced,0.9400204467773439
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,8,4,balanced,0.1400115203857422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,8,4,balanced,0.14061568260192872
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,8,4,balanced,0.1438361644744873
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,8,4,balanced,0.14644991874694824
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,8,4,balanced,0.14828160285949707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,8,4,balanced,1.2240486145019531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,8,4,balanced,0.15400704383850097
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,8,4,balanced,0.15560319900512695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,8,4,balanced,0.16214399337768554
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,8,4,balanced,0.1707084846496582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,8,4,balanced,0.18955392837524415
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,8,4,balanced,0.2064691162109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,8,4,balanced,0.245611515045166
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,8,4,balanced,1.7414770507812498
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,8,4,balanced,0.28070016860961916
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,8,4,balanced,0.3541363143920898
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,8,4,balanced,0.4305088043212891
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,8,4,power_law_1.01,0.05669248104095459
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,8,4,balanced,2.2309721374511717
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,8,4,balanced,0.6459187316894531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,8,4,power_law_1.01,0.06822527885437012
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,8,4,balanced,0.7939622497558594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,8,4,power_law_1.01,0.07475584030151368
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,8,4,power_law_1.01,0.0898521614074707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,8,4,power_law_1.01,0.12160127639770509
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,8,4,power_law_1.01,0.14190591812133788
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,8,4,balanced,1.1595980834960937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,8,4,power_law_1.01,0.18799104690551757
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,8,4,power_law_1.01,0.18484352111816407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,8,4,balanced,1.5181170654296874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,8,4,power_law_1.01,0.19187583923339843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,8,4,power_law_1.01,0.20112768173217774
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,8,4,power_law_1.01,0.20370431900024416
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,8,4,power_law_1.01,0.21700864791870117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,8,4,power_law_1.01,0.21586944580078127
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,8,4,power_law_1.01,0.2574336051940918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,8,4,power_law_1.01,0.09925888061523438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,8,4,power_law_1.01,0.12609024047851564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,8,4,power_law_1.01,0.30115072250366215
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,8,4,power_law_1.01,0.18605056762695313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,8,4,power_law_1.01,0.37259777069091793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,8,4,power_law_1.01,0.46366081237792967
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,8,4,power_law_1.01,0.6159244918823242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,8,4,power_law_1.01,0.7417984008789062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,8,4,power_law_1.01,0.2333734321594238
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,8,4,power_law_1.01,0.05402624130249023
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,8,4,power_law_1.01,1.0326489257812501
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,8,4,power_law_1.01,0.06585472106933594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,8,4,power_law_1.01,0.07064703941345216
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,8,4,power_law_1.01,0.0887168025970459
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,8,4,power_law_1.01,1.3162701416015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,8,4,power_law_1.01,0.09136128425598145
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,8,4,power_law_1.01,0.09465727806091309
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,8,4,power_law_1.01,0.11459839820861817
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,8,4,power_law_1.01,0.12111743927001953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,8,4,power_law_1.01,0.12173567771911623
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,8,4,power_law_1.01,0.037114880084991454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,8,4,power_law_1.01,0.049200639724731446
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,8,4,power_law_1.01,1.88719482421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,8,4,power_law_1.01,0.1371072006225586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,8,4,power_law_1.01,0.14889984130859374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,8,4,power_law_1.01,0.14763903617858887
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,8,4,power_law_1.01,0.15769215583801272
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,8,4,power_law_1.01,0.2203673553466797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,8,4,power_law_1.01,0.2740595245361328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,8,4,power_law_1.01,0.3131878471374512
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,8,4,power_law_1.01,2.4344677734375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,8,4,power_law_1.01,0.39878143310546876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,8,4,power_law_1.01,0.5187660980224609
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,8,4,power_law_1.01,0.1268070411682129
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,8,4,power_law_1.01,0.1404736042022705
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,8,4,power_law_1.01,0.693017578125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,8,4,power_law_1.2,0.05647488117218018
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,8,4,power_law_1.2,0.06759295940399171
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,8,4,power_law_1.01,0.18091392517089844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,8,4,power_law_1.01,0.8732518768310547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,8,4,power_law_1.2,0.10437503814697266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,8,4,power_law_1.2,0.11054464340209962
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,8,4,power_law_1.2,0.12250368118286133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,8,4,power_law_1.2,0.12808320045471192
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,8,4,power_law_1.01,1.2499148559570312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,8,4,power_law_1.2,0.18156415939331055
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,8,4,power_law_1.2,0.181209602355957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,8,4,power_law_1.2,0.18805887222290038
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,8,4,power_law_1.2,0.20795520782470703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,8,4,power_law_1.2,0.21434112548828127
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,8,4,power_law_1.01,1.661274871826172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,8,4,power_law_1.2,0.22069631576538087
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,8,4,power_law_1.2,0.24057344436645506
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,8,4,power_law_1.2,0.27202688217163085
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,8,4,power_law_1.2,0.08399231910705567
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,8,4,power_law_1.2,0.32433792114257814
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,8,4,power_law_1.2,0.3940198516845703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,8,4,power_law_1.2,0.46933631896972655
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,8,4,power_law_1.2,0.1954867172241211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,8,4,power_law_1.2,0.6096448135375977
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,8,4,power_law_1.2,0.20440832138061524
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,8,4,power_law_1.2,0.7603443145751954
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,8,4,power_law_1.2,0.037032959461212156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,8,4,power_law_1.2,0.04817791938781738
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,8,4,power_law_1.2,0.06702335834503173
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,8,4,power_law_1.2,0.051555838584899906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,8,4,power_law_1.2,0.061167359352111816
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,8,4,power_law_1.2,0.06769408226013184
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,8,4,power_law_1.2,0.08087167739868165
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,8,4,power_law_1.2,0.08869888305664062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,8,4,power_law_1.2,0.09400447845458984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,8,4,power_law_1.2,1.3806874084472658
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,8,4,power_law_1.2,0.11789055824279786
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,8,4,power_law_1.2,0.12171263694763183
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,8,4,power_law_1.2,0.12965120315551756
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,8,4,power_law_1.2,0.1376473617553711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,8,4,power_law_1.2,0.142871036529541
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,8,4,power_law_1.2,0.14629504203796387
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,8,4,power_law_1.2,0.15115776062011718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,8,4,power_law_1.2,1.9505369567871091
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,8,4,power_law_1.2,1.039767074584961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,8,4,power_law_1.2,0.1697920036315918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,8,4,power_law_1.2,0.19703680038452148
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,8,4,power_law_1.2,0.235284481048584
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,8,4,power_law_1.2,0.11336447715759276
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,8,4,power_law_1.2,0.2761062431335449
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,8,4,power_law_1.2,0.31762176513671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,8,4,power_law_1.2,0.4128358459472656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,8,4,power_law_1.2,2.628400573730469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,8,4,power_law_1.2,0.537264633178711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,16,2,balanced,0.02940416097640991
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,16,2,balanced,0.04973567962646484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,16,2,balanced,0.05740543842315674
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,16,2,balanced,0.07294079780578613
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,8,4,power_law_1.2,0.7163609313964844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,16,2,balanced,0.10613247871398926
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,16,2,balanced,0.16989952087402344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,16,2,balanced,0.17070207595825196
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,16,2,balanced,0.1737830352783203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,16,2,balanced,0.17288063049316407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,8,4,power_law_1.2,0.8951641845703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,16,2,balanced,0.1758438491821289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,16,2,balanced,0.18028799057006836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,16,2,balanced,0.21810943603515626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,16,2,balanced,0.22017919540405276
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,16,2,balanced,0.2233830451965332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,16,2,balanced,0.22708608627319338
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,8,4,power_law_1.2,1.2684121704101563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,16,2,balanced,0.2400499153137207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,16,2,balanced,0.2582310485839844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,16,2,balanced,0.27550079345703127
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,16,2,balanced,0.3131852722167969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,16,2,balanced,0.3537420654296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,8,4,power_law_1.2,1.7377062988281249
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,16,2,balanced,0.5747174453735352
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,16,2,balanced,0.029672958850860597
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,16,2,balanced,0.6756057739257812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,16,2,balanced,0.032632319927215575
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,16,2,balanced,0.04035967826843262
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,16,2,balanced,0.05827839851379395
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,16,2,balanced,0.23395711898803712
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,16,2,balanced,0.9695603179931641
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,16,2,balanced,0.08873727798461914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,16,2,balanced,0.14787967681884767
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,16,2,balanced,0.1484992027282715
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,16,2,balanced,0.15100031852722168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,16,2,balanced,0.15143168449401856
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,16,2,balanced,0.1532595157623291
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,16,2,balanced,1.2565350341796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,16,2,balanced,0.1555238437652588
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,16,2,balanced,0.19460351943969728
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,16,2,balanced,0.1986419105529785
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,16,2,balanced,0.20049407958984372
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,16,2,balanced,0.20319616317749026
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,16,2,balanced,0.21022464752197267
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,16,2,balanced,0.21591936111450197
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,16,2,balanced,0.23049087524414064
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,16,2,balanced,0.24457855224609376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,16,2,balanced,0.27906944274902346
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,16,2,balanced,1.7924684143066407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,16,2,balanced,0.31371135711669923
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,16,2,balanced,0.38254337310791015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,16,2,balanced,0.45024257659912104
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,16,2,power_law_1.01,0.039540479183197016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,16,2,power_law_1.01,0.05769599914550781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,16,2,balanced,2.369706268310547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,16,2,balanced,0.6982105255126954
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,16,2,power_law_1.01,0.06576640129089355
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,16,2,power_law_1.01,0.07885951995849609
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,16,2,balanced,0.8411532592773436
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,16,2,power_law_1.01,0.13737215995788574
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,16,2,power_law_1.01,0.11719039916992187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,16,2,power_law_1.01,0.15262463569641113
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,16,2,balanced,1.2311808013916017
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,16,2,power_law_1.01,0.15763456344604493
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,16,2,power_law_1.01,0.19648128509521484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,16,2,power_law_1.01,0.20370431900024416
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,16,2,power_law_1.01,0.21216768264770508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,16,2,power_law_1.01,0.2058995246887207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,16,2,balanced,1.6302694702148437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,16,2,power_law_1.01,0.2185868835449219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,16,2,power_law_1.01,0.2320627212524414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,16,2,power_law_1.01,0.25170560836791994
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,16,2,power_law_1.01,0.28073343276977536
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,16,2,power_law_1.01,0.13984640121459962
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,16,2,power_law_1.01,0.14313599586486817
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,16,2,power_law_1.01,0.4472972869873047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,16,2,power_law_1.01,0.5917657470703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,16,2,power_law_1.01,0.7295795440673828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,16,2,power_law_1.01,0.03273344039916992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,16,2,power_law_1.01,1.0159270477294922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,16,2,power_law_1.01,0.03989504098892212
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,16,2,power_law_1.01,0.09604736328124999
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,16,2,power_law_1.01,0.05266176223754883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,16,2,power_law_1.01,0.36257919311523434
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,16,2,power_law_1.01,1.2858982849121094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,16,2,power_law_1.01,0.06538879871368408
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,16,2,power_law_1.01,0.08184703826904297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,16,2,power_law_1.01,0.09862144470214844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,16,2,power_law_1.01,0.11759488105773926
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,16,2,power_law_1.01,0.11900927543640136
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,16,2,power_law_1.01,0.1267788791656494
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,16,2,power_law_1.01,0.13654144287109377
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,16,2,power_law_1.01,1.8625152587890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,16,2,power_law_1.01,0.1803443145751953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,16,2,power_law_1.01,0.17987071990966796
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,16,2,power_law_1.01,0.1871552085876465
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,16,2,power_law_1.01,0.19041791915893555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,16,2,power_law_1.01,0.19541248321533203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,16,2,power_law_1.01,0.22092031478881835
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,16,2,power_law_1.01,2.4119065856933597
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,16,2,power_law_1.01,0.25704959869384764
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,16,2,power_law_1.01,0.3152639961242676
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,16,2,power_law_1.01,0.12954112052917482
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,16,2,power_law_1.01,0.4504742431640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,16,2,power_law_1.01,0.1719219207763672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,16,2,power_law_1.01,0.5396505737304688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,16,2,power_law_1.2,0.0435916805267334
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,16,2,power_law_1.01,0.7594892883300781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,16,2,power_law_1.2,0.05767680168151855
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,16,2,power_law_1.2,0.06210048198699951
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,16,2,power_law_1.01,0.929733123779297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,16,2,power_law_1.2,0.07797247886657714
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,16,2,power_law_1.2,0.09305983543395996
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,16,2,power_law_1.2,0.1074841594696045
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,16,2,power_law_1.01,1.7729087829589845
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,16,2,power_law_1.2,0.14950400352478027
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,16,2,power_law_1.2,0.16170751571655273
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,16,2,power_law_1.01,0.35719425201416016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,16,2,power_law_1.2,0.19525888442993164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,16,2,power_law_1.2,0.13408896446228027
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,16,2,power_law_1.2,0.14332672119140627
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,16,2,power_law_1.2,0.20314239501953124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,16,2,power_law_1.2,0.20904191970825195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,16,2,power_law_1.01,1.3412736511230468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,16,2,power_law_1.2,0.2139967918395996
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,16,2,power_law_1.2,0.22185087203979492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,16,2,power_law_1.2,0.23710336685180664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,16,2,power_law_1.2,0.2605286407470703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,16,2,power_law_1.2,0.2904447937011719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,16,2,power_law_1.2,0.38640766143798827
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,16,2,power_law_1.2,0.4573004913330078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,16,2,power_law_1.2,0.6120627212524414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,16,2,power_law_1.2,0.03271807909011841
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,16,2,power_law_1.2,0.13948927879333498
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,16,2,power_law_1.2,0.04009600162506104
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,16,2,power_law_1.2,0.048244481086730955
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,16,2,power_law_1.2,1.0558604431152343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,16,2,power_law_1.2,0.061538558006286624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,16,2,power_law_1.2,0.07589375972747803
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,16,2,power_law_1.2,1.3393075561523438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,16,2,power_law_1.2,0.08869119644165038
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,16,2,power_law_1.2,0.11742848396301268
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,16,2,power_law_1.2,0.11884544372558595
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,16,2,power_law_1.2,0.12289152145385743
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,16,2,power_law_1.2,0.1255244827270508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,16,2,power_law_1.2,0.13456255912780762
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,16,2,power_law_1.2,0.1705727958679199
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,16,2,power_law_1.2,0.1813337516784668
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,16,2,power_law_1.2,0.7427212524414062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,16,2,power_law_1.2,0.18336896896362304
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,16,2,power_law_1.2,1.9493746948242188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,16,2,power_law_1.2,0.1867788887023926
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,16,2,power_law_1.2,0.19112831115722656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,16,2,power_law_1.2,0.20993024826049805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,16,2,power_law_1.2,0.23950336456298826
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,16,2,power_law_1.2,0.27587968826293946
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,16,2,power_law_1.2,0.3239347076416016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,16,2,power_law_1.2,0.3687615966796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,16,2,power_law_1.2,0.46353023529052734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,16,2,power_law_1.2,2.4853619384765624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,32,1,balanced,0.03661312103271484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,16,2,power_law_1.2,0.5570380783081055
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,32,1,balanced,0.06065152168273926
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,32,1,balanced,0.07058752059936524
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,32,1,balanced,0.08800959587097168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,16,2,power_law_1.2,0.7693209838867188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,32,1,balanced,0.12772480010986328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,32,1,balanced,0.2199564743041992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,32,1,balanced,0.21838336944580078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,16,2,power_law_1.2,0.9685874938964844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,32,1,balanced,0.21754623413085938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,32,1,balanced,0.21985855102539062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,32,1,balanced,0.22150720596313475
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,32,1,balanced,0.22533952713012698
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,32,1,balanced,0.23179456710815433
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,32,1,balanced,0.24917823791503907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,32,1,balanced,0.23937599182128907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,32,1,balanced,0.2568620872497559
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,32,1,balanced,0.2609996795654297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,32,1,balanced,0.26669696807861326
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,32,1,balanced,0.2842419242858887
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,32,1,balanced,0.2978988838195801
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,32,1,balanced,0.344007682800293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,16,2,power_law_1.2,1.809537353515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,32,1,balanced,0.3906412887573242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,32,1,balanced,0.6241190338134766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,32,1,balanced,0.03945472002029419
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,16,2,power_law_1.2,1.3735026550292968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,32,1,balanced,0.04472447872161865
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,32,1,balanced,0.941107177734375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,32,1,balanced,0.05811583995819092
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,32,1,balanced,0.0857151985168457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,32,1,balanced,0.13256896018981934
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,32,1,balanced,0.23329599380493163
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,32,1,balanced,1.259491195678711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,32,1,balanced,0.233319034576416
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,32,1,balanced,0.23597248077392577
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,32,1,balanced,0.6759641265869141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,32,1,balanced,0.23716928482055666
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,32,1,balanced,0.23853439331054688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,32,1,balanced,0.24464319229125975
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,32,1,balanced,0.24508544921875003
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,32,1,balanced,0.25017984390258785
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,32,1,balanced,1.8808499145507813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,32,1,balanced,0.326416015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,32,1,balanced,0.33127422332763673
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,32,1,balanced,0.3345612716674805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,32,1,balanced,0.3494188690185547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,32,1,balanced,0.3612665557861328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,32,1,balanced,0.3923814392089844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,32,1,balanced,0.42451713562011717
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,32,1,balanced,2.4792807006835935
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,32,1,balanced,0.23982976913452148
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,32,1,balanced,0.4908448028564453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,32,1,balanced,0.553045768737793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,32,1,power_law_1.01,0.03664448022842407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,32,1,power_law_1.01,0.05488063812255859
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,32,1,power_law_1.01,0.07034111976623535
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,32,1,power_law_1.01,0.08946304321289063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,32,1,power_law_1.01,0.11183103561401367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,32,1,power_law_1.01,0.14621631622314452
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,32,1,balanced,1.0482374572753907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,32,1,power_law_1.01,0.17999744415283203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,32,1,power_law_1.01,0.1846143913269043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,32,1,power_law_1.01,0.18894079208374023
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,32,1,power_law_1.01,0.19099584579467774
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,32,1,power_law_1.01,0.20510528564453123
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,32,1,power_law_1.01,0.2151385688781738
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,32,1,power_law_1.01,0.22429439544677737
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,32,1,power_law_1.01,0.2461734390258789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,32,1,power_law_1.01,0.23762752532958986
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,32,1,balanced,0.9132978820800781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,32,1,power_law_1.01,0.2430201530456543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,32,1,power_law_1.01,0.25729536056518554
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,32,1,power_law_1.01,0.28038848876953126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,32,1,power_law_1.01,0.3069254493713379
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,32,1,balanced,2.0616268920898437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,32,1,power_law_1.01,0.3638131332397461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,32,1,power_law_1.01,0.44020416259765627
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,32,1,balanced,1.5540223693847657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,32,1,power_law_1.01,0.6082841491699218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,32,1,power_law_1.01,0.03916800022125244
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,32,1,power_law_1.01,0.7248537445068359
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,32,1,power_law_1.01,0.04447103977203369
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,32,1,power_law_1.01,0.05985599994659424
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,32,1,power_law_1.01,0.0850476837158203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,32,1,power_law_1.01,1.00565185546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,32,1,power_law_1.01,0.1120019245147705
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,32,1,power_law_1.01,0.14100799560546876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,32,1,power_law_1.01,0.18369279861450197
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,32,1,power_law_1.01,1.3049440002441406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,32,1,power_law_1.01,0.19048511505126953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,32,1,power_law_1.01,0.19494144439697264
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,32,1,power_law_1.01,0.1983692741394043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,32,1,power_law_1.01,0.2062073516845703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,32,1,power_law_1.01,0.2145132827758789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,32,1,power_law_1.01,0.22256383895874024
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,32,1,power_law_1.01,0.29838783264160157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,32,1,power_law_1.01,0.30524032592773437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,32,1,power_law_1.01,1.889637145996094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,32,1,power_law_1.01,0.31583744049072265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,32,1,power_law_1.01,0.3387974548339844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,32,1,power_law_1.01,0.3725446319580078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,32,1,power_law_1.01,0.4459872055053711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,32,1,power_law_1.01,0.5007468795776366
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,32,1,power_law_1.01,2.466229705810547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,32,1,power_law_1.01,0.21694143295288085
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,32,1,power_law_1.01,0.6098214340209961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,32,1,power_law_1.2,0.03753024101257324
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,32,1,power_law_1.01,0.7462700653076172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,32,1,power_law_1.2,0.058989439010620114
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,32,1,power_law_1.2,0.07043647766113281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,32,1,power_law_1.2,0.0854355239868164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,32,1,power_law_1.01,0.9910214233398438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,32,1,power_law_1.2,0.10739392280578612
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,32,1,power_law_1.2,0.14038335800170898
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,32,1,power_law_1.2,0.17836416244506836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,32,1,power_law_1.2,0.18347776412963868
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,32,1,power_law_1.01,1.2500838470458984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,32,1,power_law_1.2,0.1920876884460449
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,32,1,power_law_1.2,0.2046112060546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,32,1,power_law_1.2,0.2137619209289551
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,32,1,power_law_1.2,0.22700735092163088
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,32,1,power_law_1.2,0.2505862426757813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,32,1,power_law_1.2,0.2413395118713379
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,32,1,power_law_1.2,0.24910015106201172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,32,1,power_law_1.01,1.742830047607422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,32,1,power_law_1.2,0.2598572731018066
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,32,1,power_law_1.2,0.2899321556091309
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,32,1,power_law_1.2,0.31349119186401364
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,32,1,power_law_1.2,0.3759481430053711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,32,1,power_law_1.01,2.2489459228515623
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,32,1,power_law_1.2,0.451464958190918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,32,1,power_law_1.2,0.6226591873168945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,32,1,power_law_1.2,0.7391910552978516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,32,1,power_law_1.2,0.039485440254211426
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,32,1,power_law_1.2,0.18778047561645508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,32,1,power_law_1.2,1.021878433227539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,32,1,power_law_1.2,0.04430975914001465
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,32,1,power_law_1.2,0.05992576122283936
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,32,1,power_law_1.2,0.08083328247070312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,32,1,power_law_1.2,0.1052889633178711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,32,1,power_law_1.2,0.13510527610778808
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,32,1,power_law_1.2,0.17976768493652345
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,32,1,power_law_1.2,0.18731647491455078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,32,1,power_law_1.2,0.19127103805541992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,32,1,power_law_1.2,0.19871871948242187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,32,1,power_law_1.2,0.20410240173339841
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,32,1,power_law_1.2,0.21150527954101564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,32,1,power_law_1.2,1.9122195434570313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,32,1,power_law_1.2,0.21718784332275393
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,32,1,power_law_1.2,0.2261952018737793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,32,1,power_law_1.2,1.3205810546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,32,1,power_law_1.2,0.3024812889099121
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,32,1,power_law_1.2,0.31207487106323245
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,32,1,power_law_1.2,0.32360000610351564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,32,1,power_law_1.2,0.36012542724609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,32,1,power_law_1.2,2.488306579589844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,32,1,power_law_1.2,0.40724094390869137
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,32,1,power_law_1.2,0.46414463043212895
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,1,32,balanced,0.0181388795375824
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,1,32,balanced,0.018170880079269407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,32,1,power_law_1.2,0.6303596878051758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,1,32,balanced,0.017957119941711425
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,32,1,power_law_1.2,0.7678534698486328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,1,32,balanced,0.018388479948043823
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,1,32,balanced,0.022785279750823974
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,1,32,balanced,0.03837951898574829
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,1,32,balanced,0.036614398956298824
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,1,32,balanced,0.040746240615844725
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,32,1,power_law_1.2,1.0099788665771485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,1,32,balanced,0.03577600002288818
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,1,32,balanced,0.040943360328674315
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,1,32,balanced,0.04089087963104248
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,1,32,balanced,0.03879040002822876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,1,32,balanced,0.046995201110839845
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,1,32,balanced,0.04660736083984375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,32,1,power_law_1.2,0.5137164688110352
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,1,32,balanced,0.047489280700683596
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,1,32,balanced,0.04749824047088623
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,32,1,power_law_1.2,1.270505599975586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,1,32,balanced,0.05348095893859863
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,1,32,balanced,0.057346558570861815
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,1,32,balanced,0.06768511772155762
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,1,32,balanced,0.09251328468322753
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,1,32,balanced,0.13005184173583983
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,1,32,balanced,0.16044288635253906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,1,32,balanced,0.03457279920578003
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,1,32,balanced,0.23976959228515624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,1,32,balanced,0.021087999343872073
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,32,1,power_law_1.2,1.7660595703125002
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,1,32,balanced,0.29896448135375975
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,1,32,balanced,0.02068351984024048
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,1,32,balanced,0.021184000968933105
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,1,32,balanced,0.021116158962249755
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,1,32,balanced,0.031508479118347164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,1,32,balanced,0.02170880079269409
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,1,32,balanced,0.032471039295196534
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,1,32,balanced,0.0325977611541748
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,1,32,balanced,0.03298943996429443
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,1,32,balanced,0.03562367916107177
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,1,32,balanced,0.03396352052688599
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,1,32,balanced,0.034588160514831545
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,1,32,balanced,0.03569407939910889
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,1,32,balanced,0.03704319953918457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,1,32,balanced,0.06305280208587646
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,1,32,balanced,0.03825664043426514
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,1,32,balanced,0.034777600765228275
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,32,1,power_law_1.2,2.2709747314453126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,1,32,balanced,0.08107775688171387
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,1,32,balanced,0.04261119842529297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,1,32,balanced,0.048198399543762205
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,1,32,balanced,0.05443327903747559
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,1,32,balanced,0.0645363187789917
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,1,32,balanced,0.10237824440002441
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,1,32,balanced,0.1281062412261963
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,1,32,balanced,0.18642431259155273
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,1,32,balanced,0.23072895050048828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,1,32,power_law_1.01,0.023630080223083497
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,1,32,power_law_1.01,0.026085119247436523
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,1,32,power_law_1.01,0.03655424118041992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,1,32,balanced,0.03183871984481811
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,1,32,power_law_1.01,0.036686079502105715
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,1,32,power_law_1.01,0.038365440368652345
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,1,32,power_law_1.01,0.04308351993560791
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,1,32,power_law_1.01,0.04154880046844482
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,1,32,power_law_1.01,0.0425600004196167
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,1,32,power_law_1.01,0.04498432159423828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,1,32,power_law_1.01,0.049492478370666504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,1,32,power_law_1.01,0.048728318214416505
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,1,32,power_law_1.01,0.05282048225402832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,1,32,power_law_1.01,0.05664512157440186
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,1,32,power_law_1.01,0.05293951988220215
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,1,32,power_law_1.01,0.05572864055633545
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,1,32,power_law_1.01,0.06015615940093995
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,1,32,power_law_1.01,0.07169663906097412
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,1,32,balanced,0.07696256160736084
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,1,32,power_law_1.01,0.09049856185913085
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,1,32,power_law_1.01,0.020052480697631835
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,1,32,power_law_1.01,0.13156352043151856
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,1,32,power_law_1.01,0.1890355110168457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,1,32,power_law_1.01,0.24702463150024415
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,1,32,power_law_1.01,0.3163507270812988
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,1,32,power_law_1.01,0.021094400882720948
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,1,32,power_law_1.01,0.45937152862548825
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,1,32,power_law_1.01,0.021584639549255373
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,1,32,balanced,0.03809792041778565
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,1,32,power_law_1.01,0.03142911911010742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,1,32,power_law_1.01,0.7899954986572266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,1,32,power_law_1.01,0.031650559902191164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,1,32,power_law_1.01,0.07458943843841552
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,1,32,power_law_1.01,0.03220223903656006
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,1,32,power_law_1.01,0.03333120107650757
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,1,32,power_law_1.01,0.9083187103271484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,1,32,power_law_1.01,0.034155519008636476
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,1,32,power_law_1.01,0.03393408060073853
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,1,32,power_law_1.01,0.034937601089477535
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,1,32,power_law_1.01,0.03613568067550659
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,1,32,power_law_1.01,0.03722624063491821
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,1,32,power_law_1.01,0.03956480026245117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,1,32,power_law_1.01,0.03986047983169556
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,1,32,power_law_1.01,0.042894082069396974
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,1,32,power_law_1.01,0.04797056198120117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,1,32,power_law_1.01,0.053349118232727054
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,1,32,power_law_1.01,0.05520512104034424
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,1,32,power_law_1.01,0.07095808029174805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,1,32,power_law_1.01,0.09752703666687011
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,1,32,power_law_1.01,0.14303104400634764
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,1,32,power_law_1.01,0.17753215789794924
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,1,32,power_law_1.01,0.021396479606628417
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,1,32,power_law_1.01,0.2381849670410156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,1,32,power_law_1.2,0.020188159942626953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,1,32,power_law_1.2,0.022873599529266358
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,1,32,power_law_1.2,0.022571520805358885
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,1,32,power_law_1.01,0.46272129058837896
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,1,32,power_law_1.2,0.03373951911926269
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,1,32,power_law_1.2,0.03758975982666016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,1,32,power_law_1.2,0.03478784084320068
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,1,32,power_law_1.2,0.03991039991378784
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,1,32,power_law_1.01,0.03702143907546997
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,1,32,power_law_1.2,0.044846081733703615
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,1,32,power_law_1.01,0.64814208984375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,1,32,power_law_1.2,0.04535039901733399
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,1,32,power_law_1.2,0.05114240169525146
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,1,32,power_law_1.2,0.055379199981689456
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,1,32,power_law_1.2,0.053265919685363775
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,1,32,power_law_1.2,0.05795199871063232
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,1,32,power_law_1.2,0.0628275203704834
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,1,32,power_law_1.2,0.07480192184448242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,1,32,power_law_1.2,0.09031680107116699
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,1,32,power_law_1.2,0.12473600387573243
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,1,32,power_law_1.2,0.1554854393005371
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,1,32,power_law_1.01,0.31958784103393556
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,1,32,power_law_1.2,0.22685056686401367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,1,32,power_law_1.2,0.307640323638916
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,1,32,power_law_1.2,0.021269760131835937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,1,32,power_law_1.2,0.043843841552734374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,1,32,power_law_1.2,0.05330431938171387
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,1,32,power_law_1.2,0.021214718818664553
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,1,32,power_law_1.2,0.021497600078582764
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,1,32,power_law_1.2,0.031357440948486325
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,1,32,power_law_1.2,0.031807999610900875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,1,32,power_law_1.2,0.4390438461303711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,1,32,power_law_1.2,0.03413503885269165
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,1,32,power_law_1.2,0.033808639049530034
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,1,32,power_law_1.2,0.03429120063781738
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,1,32,power_law_1.2,0.03472127914428711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,1,32,power_law_1.2,0.03772160053253174
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,1,32,power_law_1.2,0.040028162002563476
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,1,32,power_law_1.2,0.03883647918701172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,1,32,power_law_1.2,0.038937599658966066
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,1,32,power_law_1.2,0.04098559856414795
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,1,32,power_law_1.2,0.04342656135559082
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,1,32,power_law_1.2,0.05411968231201172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,1,32,power_law_1.2,0.048428797721862794
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,1,32,power_law_1.2,0.056789760589599614
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,1,32,power_law_1.2,0.06649216175079345
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,1,32,power_law_1.2,0.08900223731994629
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,1,32,power_law_1.2,0.03245055913925171
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,1,32,power_law_1.2,0.10980095863342285
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,1,32,power_law_1.2,0.17212160110473634
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,1,32,power_law_1.2,0.9336819458007813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,1,32,power_law_1.2,0.5736678314208985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,2,16,balanced,0.017103359699249268
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,2,16,balanced,0.016643840074539184
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,2,16,balanced,0.017564159631729127
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,2,16,balanced,0.017950719594955443
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,1,32,power_law_1.2,0.3396211242675781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,2,16,balanced,0.023719680309295655
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,2,16,balanced,0.03674623966217041
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,2,16,balanced,0.031215360164642336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,2,16,balanced,0.039137279987335204
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,2,16,balanced,0.036942079067230224
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,2,16,balanced,0.037576959133148194
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,2,16,balanced,0.038380799293518064
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,1,32,power_law_1.2,0.6606156921386719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,1,32,power_law_1.2,0.42406654357910156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,2,16,balanced,0.04188416004180908
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,2,16,balanced,0.04277120113372802
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,2,16,balanced,0.04051199913024902
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,2,16,balanced,0.04674943923950196
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,2,16,balanced,0.04894336223602295
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,2,16,balanced,0.05236991882324219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,2,16,balanced,0.05691135883331298
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,2,16,balanced,0.06173952102661133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,2,16,balanced,0.07344895839691162
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,2,16,balanced,0.08173184394836426
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,2,16,balanced,0.10492544174194336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,2,16,balanced,0.12558591842651368
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,2,16,balanced,0.1776358413696289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,2,16,balanced,0.22445823669433596
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,1,32,power_law_1.2,0.21752832412719725
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,2,16,balanced,0.020030720233917238
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,2,16,balanced,0.3327347183227539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,2,16,balanced,0.0199782395362854
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,2,16,balanced,0.02041856050491333
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,2,16,balanced,0.42229888916015623
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,2,16,balanced,0.020590078830718995
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,2,16,balanced,0.022163200378417968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,2,16,balanced,0.030420479774475095
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,2,16,balanced,0.03371520042419433
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,2,16,balanced,0.033873920440673826
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,2,16,balanced,0.03504512071609497
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,2,16,balanced,0.03619328022003174
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,2,16,balanced,0.03265664100646973
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,2,16,balanced,0.03524096012115478
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,2,16,balanced,0.0377292799949646
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,2,16,balanced,0.03940736055374146
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,2,16,balanced,0.043022079467773436
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,2,16,balanced,0.04866176128387451
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,2,16,balanced,0.05906688213348389
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,2,16,balanced,0.0689408016204834
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,2,16,balanced,0.09364480018615723
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,2,16,balanced,0.11358207702636718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,2,16,balanced,0.15449472427368166
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,2,16,balanced,0.19288320541381837
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,2,16,balanced,0.03136512041091919
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,2,16,balanced,0.032119040489196775
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,2,16,power_law_1.01,0.019852800369262697
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,2,16,power_law_1.01,0.022914559841156007
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,2,16,balanced,0.36540672302246097
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,2,16,power_law_1.01,0.02211456060409546
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,2,16,power_law_1.01,0.02532480001449585
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,2,16,power_law_1.01,0.0339737606048584
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,2,16,power_law_1.01,0.03435391902923584
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,2,16,power_law_1.01,0.03580415964126587
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,2,16,power_law_1.01,0.04146944046020508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,2,16,power_law_1.01,0.04062079906463623
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,2,16,power_law_1.01,0.039468801021575926
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,2,16,power_law_1.01,0.046369280815124515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,2,16,power_law_1.01,0.049135360717773434
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,2,16,power_law_1.01,0.04938240051269531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,2,16,power_law_1.01,0.04747392177581787
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,2,16,power_law_1.01,0.04973696231842041
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,2,16,power_law_1.01,0.05230463981628418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,1,32,power_law_1.2,0.810662384033203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,2,16,power_law_1.01,0.06583040237426759
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,2,16,power_law_1.01,0.07604735851287842
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,2,16,power_law_1.01,0.08897919654846191
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,2,16,power_law_1.01,0.11562368392944336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,2,16,power_law_1.01,0.16404224395751954
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,2,16,power_law_1.01,0.1799500846862793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,2,16,power_law_1.01,0.2944256019592285
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,2,16,power_law_1.01,0.3864422225952148
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,2,16,power_law_1.01,0.020582399368286132
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,2,16,power_law_1.01,0.5581158447265625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,2,16,power_law_1.01,0.02123647928237915
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,2,16,power_law_1.01,0.02192768096923828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,2,16,balanced,0.03279360055923462
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,2,16,power_law_1.01,0.7466226959228516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,2,16,power_law_1.01,0.02187391996383667
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,2,16,power_law_1.01,0.030361599922180176
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,2,16,power_law_1.01,0.031819519996643064
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,2,16,power_law_1.01,0.03222527980804443
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,2,16,power_law_1.01,0.033369600772857666
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,2,16,power_law_1.01,0.035741438865661626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,2,16,power_law_1.01,0.03654911994934082
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,2,16,power_law_1.01,0.036445438861846924
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,2,16,power_law_1.01,0.037328639030456544
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,2,16,power_law_1.01,0.038818559646606444
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,2,16,power_law_1.01,0.040547838211059575
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,2,16,power_law_1.01,0.04510335922241211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,2,16,power_law_1.01,0.05119232177734375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,2,16,power_law_1.01,0.06244351863861084
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,2,16,power_law_1.01,0.05617152214050293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,2,16,balanced,0.2879347229003906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,2,16,power_law_1.01,0.07776000022888183
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,2,16,power_law_1.01,0.09543807983398438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,2,16,power_law_1.01,0.13352959632873534
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,2,16,power_law_1.01,0.03382400035858154
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,2,16,power_law_1.01,0.15438207626342773
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,2,16,power_law_1.01,0.23628288269042969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,2,16,power_law_1.2,0.022685439586639405
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,2,16,power_law_1.01,0.36391166687011717
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,2,16,power_law_1.2,0.024131839275360105
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,2,16,power_law_1.01,0.4246195220947266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,2,16,power_law_1.2,0.02387712001800537
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,2,16,power_law_1.2,0.023379199504852295
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,2,16,power_law_1.2,0.025525760650634766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,2,16,power_law_1.01,0.5733567810058593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,2,16,power_law_1.2,0.03669503927230835
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,2,16,power_law_1.2,0.040677118301391604
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,2,16,power_law_1.2,0.04027775764465332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,2,16,power_law_1.2,0.04635263919830322
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,2,16,power_law_1.2,0.041912322044372556
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,2,16,power_law_1.2,0.046936321258544925
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,2,16,power_law_1.2,0.05169280052185059
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,2,16,power_law_1.2,0.048473601341247556
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,2,16,power_law_1.2,0.05118720054626465
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,2,16,power_law_1.2,0.05404799938201904
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,2,16,power_law_1.2,0.059128317832946774
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,2,16,power_law_1.2,0.069235200881958
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,2,16,power_law_1.2,0.08103039741516113
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,2,16,power_law_1.2,0.10156160354614259
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,2,16,power_law_1.2,0.13109631538391114
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,2,16,power_law_1.2,0.1733593559265137
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,2,16,power_law_1.2,0.03919487953186035
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,2,16,power_law_1.2,0.3672972869873047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,2,16,power_law_1.2,0.24555776596069334
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,2,16,power_law_1.2,0.02065279960632324
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,2,16,power_law_1.01,0.03215872049331665
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,2,16,power_law_1.2,0.020651519298553467
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,2,16,power_law_1.2,0.02159872055053711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,2,16,power_law_1.2,0.021391360759735106
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,2,16,power_law_1.2,0.031103999614715577
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,2,16,power_law_1.2,0.03108223915100098
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,2,16,power_law_1.2,0.03201280117034912
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,2,16,power_law_1.2,0.032590079307556155
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,2,16,power_law_1.2,0.03319808006286621
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,2,16,power_law_1.2,0.0341811203956604
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,2,16,power_law_1.2,0.03631360054016113
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,2,16,power_law_1.2,0.03875200033187866
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,2,16,power_law_1.2,0.03701888084411621
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,2,16,power_law_1.2,0.03816191911697388
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,2,16,power_law_1.2,0.04013184070587158
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,2,16,power_law_1.2,0.0425548791885376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,2,16,power_law_1.2,0.04620543956756592
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,2,16,power_law_1.2,0.055278081893920895
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,2,16,power_law_1.2,0.06553343772888184
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,2,16,power_law_1.2,0.07826047897338867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,2,16,power_law_1.2,0.10033920288085937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,2,16,power_law_1.2,0.15347328186035156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,2,16,power_law_1.2,0.1907481575012207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,2,16,power_law_1.2,0.6261836624145507
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,4,8,balanced,0.016925439834594727
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,4,8,balanced,0.017210880517959597
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,1,32,power_law_1.2,1.2271129608154296
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,4,8,balanced,0.01812351942062378
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,4,8,balanced,0.02439807891845703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,4,8,balanced,0.01961727976799011
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,4,8,balanced,0.039870719909667965
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,4,8,balanced,0.033593599796295164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,2,16,power_law_1.2,0.38483711242675783
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,4,8,balanced,0.04137087821960449
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,4,8,balanced,0.04005760192871094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,4,8,balanced,0.044144639968872065
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,4,8,balanced,0.042154240608215335
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,4,8,balanced,0.04490367889404297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,4,8,balanced,0.04685311794281006
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,4,8,balanced,0.05035136222839356
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,4,8,balanced,0.052700161933898926
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,2,16,power_law_1.2,0.4323763275146485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,4,8,balanced,0.05475071907043457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,4,8,balanced,0.059333119392395016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,4,8,balanced,0.06882815837860108
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,4,8,balanced,0.07774079799652099
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,4,8,balanced,0.09761792182922363
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,4,8,balanced,0.11557375907897949
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,2,16,power_law_1.2,0.28451072692871093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,4,8,balanced,0.19210111618041992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,4,8,balanced,0.27399168014526365
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,4,8,balanced,0.3543129730224609
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,4,8,balanced,0.020098559856414795
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,4,8,balanced,0.5265740966796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,4,8,balanced,0.020193281173706053
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,4,8,balanced,0.021095681190490722
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,4,8,balanced,0.023576319217681885
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,4,8,balanced,0.035743999481201175
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,4,8,balanced,0.6814579010009766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,4,8,balanced,0.03485824108123779
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,4,8,balanced,0.03669120073318481
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,4,8,balanced,0.03812608003616333
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,4,8,balanced,0.04015872001647949
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,4,8,balanced,0.03720063924789428
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,4,8,balanced,0.15595775604248047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,4,8,balanced,0.03709696054458618
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,4,8,balanced,0.04054399967193604
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,4,8,balanced,0.042213120460510255
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,4,8,balanced,0.043196158409118654
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,4,8,balanced,0.044994559288024906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,4,8,balanced,0.04844287872314453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,4,8,balanced,0.05873792171478272
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,4,8,balanced,0.06682752132415772
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,4,8,balanced,0.08870271682739259
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,4,8,balanced,0.107258882522583
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,4,8,balanced,0.1464575958251953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,4,8,balanced,0.021459200382232667
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,4,8,balanced,0.18076799392700196
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,4,8,balanced,0.25955455780029296
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,4,8,balanced,0.330772476196289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,2,16,power_law_1.2,0.563125762939453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,4,8,power_law_1.01,0.018108160495758058
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,4,8,balanced,0.4909312057495117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,4,8,power_law_1.01,0.023047680854797366
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,4,8,power_law_1.01,0.024107520580291746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,4,8,power_law_1.01,0.02636672019958496
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,4,8,power_law_1.01,0.026624000072479247
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,4,8,power_law_1.01,0.035723519325256345
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,4,8,power_law_1.01,0.03806848049163818
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,4,8,power_law_1.01,0.03991296052932739
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,4,8,power_law_1.01,0.038483200073242185
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,4,8,power_law_1.01,0.043610877990722656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,4,8,power_law_1.01,0.045887999534606934
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,4,8,power_law_1.01,0.0470579195022583
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,4,8,power_law_1.01,0.04641151905059814
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,4,8,power_law_1.01,0.047744002342224125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,4,8,power_law_1.01,0.04890751838684082
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,4,8,power_law_1.01,0.05073919773101807
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,4,8,power_law_1.01,0.05627520084381103
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,4,8,power_law_1.01,0.06504576206207276
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,4,8,power_law_1.01,0.078090238571167
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,4,8,power_law_1.01,0.09661312103271484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,4,8,power_law_1.01,0.10650624275207518
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,4,8,power_law_1.01,0.1503001594543457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,4,8,power_law_1.01,0.1838092803955078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,4,8,balanced,0.6412480163574219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,4,8,power_law_1.01,0.288014087677002
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,4,8,power_law_1.01,0.35637889862060546
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,4,8,power_law_1.01,0.020673279762268064
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,4,8,power_law_1.01,0.02142080068588257
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,4,8,power_law_1.01,0.022279679775238037
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,4,8,power_law_1.01,0.5049433517456055
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,4,8,power_law_1.01,0.022464001178741456
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,4,8,power_law_1.01,0.02320768117904663
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,4,8,power_law_1.01,0.03417727947235107
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,4,8,power_law_1.01,0.0348633599281311
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,4,8,power_law_1.01,0.6500313568115235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,4,8,power_law_1.01,0.03727231979370117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,4,8,power_law_1.01,0.0375436806678772
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,4,8,power_law_1.01,0.04150015830993652
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,4,8,power_law_1.01,0.03681792020797729
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,4,8,power_law_1.01,0.04161920070648194
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,4,8,power_law_1.01,0.03808511972427368
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,4,8,power_law_1.01,0.04261631965637207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,4,8,power_law_1.01,0.045506558418273925
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,4,8,power_law_1.01,0.048592638969421384
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,4,8,power_law_1.01,0.05586175918579102
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,4,8,power_law_1.01,0.06660736083984375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,4,8,power_law_1.01,0.08531711578369142
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,4,8,power_law_1.01,0.10994815826416016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,4,8,power_law_1.01,0.14485247611999513
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,4,8,power_law_1.01,0.19548288345336914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,4,8,power_law_1.01,0.2536537551879883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,4,8,power_law_1.01,0.0323417592048645
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,4,8,power_law_1.01,0.3462284851074219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,4,8,power_law_1.2,0.018579200506210328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,4,8,power_law_1.2,0.022042880058288576
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,4,8,power_law_1.01,0.5364633560180664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,4,8,power_law_1.2,0.02473344087600708
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,4,8,power_law_1.2,0.028491520881652833
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,4,8,power_law_1.2,0.03498111963272095
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,4,8,power_law_1.01,0.6415449523925781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,4,8,power_law_1.2,0.037876479625701905
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,4,8,power_law_1.2,0.04024191856384277
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,4,8,power_law_1.2,0.04456064224243164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,4,8,power_law_1.2,0.045715198516845704
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,4,8,power_law_1.2,0.04662015914916992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,4,8,power_law_1.2,0.04693247795104981
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,4,8,power_law_1.2,0.05007359981536865
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,4,8,power_law_1.2,0.05237760066986084
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,4,8,power_law_1.2,0.05849472045898437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,4,8,power_law_1.2,0.0660864019393921
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,4,8,power_law_1.2,0.01794816017150879
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,4,8,power_law_1.2,0.07792384147644042
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,4,8,power_law_1.2,0.09925888061523438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,4,8,power_law_1.2,0.11596416473388671
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,4,8,power_law_1.2,0.16224767684936522
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,2,16,power_law_1.2,0.7615987396240234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,4,8,power_law_1.2,0.03992831945419312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,4,8,power_law_1.2,0.2062118339538574
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,4,8,power_law_1.2,0.04644608020782471
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,4,8,power_law_1.2,0.020721919536590576
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,4,8,power_law_1.2,0.021267199516296388
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,4,8,power_law_1.2,0.0220032000541687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,4,8,power_law_1.2,0.022405118942260743
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,4,8,power_law_1.2,0.3668646240234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,4,8,power_law_1.2,0.03442816019058227
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,4,8,power_law_1.2,0.03428607940673828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,4,8,power_law_1.2,0.035770881175994876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,4,8,power_law_1.2,0.036693758964538574
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,4,8,power_law_1.2,0.03810175895690918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,4,8,power_law_1.2,0.04225535869598389
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,4,8,power_law_1.2,0.03868160009384155
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,4,8,power_law_1.2,0.04028031826019287
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,4,8,power_law_1.2,0.3005273628234863
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,4,8,power_law_1.2,0.04229887962341309
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,4,8,power_law_1.2,0.04454271793365479
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,4,8,power_law_1.2,0.04457471847534179
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,4,8,power_law_1.2,0.048890881538391114
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,4,8,power_law_1.2,0.05961599826812745
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,4,8,power_law_1.2,0.06840320110321045
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,4,8,power_law_1.2,0.08846464157104492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,4,8,power_law_1.2,0.11222271919250489
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,4,8,power_law_1.2,0.15536000251770019
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,4,8,power_law_1.2,0.2055232048034668
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,4,8,power_law_1.2,0.02331775903701782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,1,32,balanced,0.03318527936935425
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,1,32,balanced,0.033850879669189454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,1,32,balanced,0.027173120975494385
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,1,32,balanced,0.04917119979858399
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,1,32,balanced,0.059991040229797364
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,1,32,balanced,0.09590784072875977
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,2,16,power_law_1.2,0.85494140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,1,32,balanced,0.09693568229675294
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,1,32,balanced,0.09725055694580079
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,4,8,power_law_1.2,0.5931967926025391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,1,32,balanced,0.0973529624938965
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,1,32,balanced,0.09793536186218262
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,1,32,balanced,0.09877632141113282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,1,32,balanced,0.10367487907409667
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,1,32,balanced,0.10561535835266114
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,1,32,balanced,0.11037055969238281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,4,8,power_law_1.2,0.28071807861328124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,1,32,balanced,0.11147775650024414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,1,32,balanced,0.11330304145812989
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,1,32,balanced,0.11011327743530272
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,1,32,balanced,0.1158118438720703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,1,32,balanced,0.12181247711181639
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,1,32,balanced,0.14113408088684082
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,1,32,balanced,0.1426534366607666
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,1,32,balanced,0.20401536941528323
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,1,32,balanced,0.24231168746948245
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,1,32,balanced,0.3396416091918945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,1,32,balanced,0.42105728149414057
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,1,32,balanced,0.026990079879760744
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,1,32,balanced,0.026849279403686522
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,1,32,balanced,0.02684416055679321
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,1,32,balanced,0.6122521591186524
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,1,32,balanced,0.029031679630279538
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,1,32,balanced,0.04190591812133789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,4,8,power_law_1.2,0.360555534362793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,1,32,balanced,0.7950592041015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,1,32,balanced,0.04993535995483399
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,1,32,balanced,0.052043519020080566
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,1,32,balanced,0.05274240016937256
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,1,32,balanced,0.05431680202484131
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,1,32,balanced,0.054200320243835455
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,1,32,balanced,0.05560575962066651
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,1,32,balanced,0.05619584083557129
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,1,32,balanced,0.06074495792388916
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,1,32,balanced,0.06506624221801757
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,1,32,balanced,0.05850240230560303
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,1,32,balanced,0.06479487895965577
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,1,32,balanced,0.07008512020111084
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,1,32,balanced,0.07966207981109619
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,1,32,balanced,0.08984959602355957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,1,32,balanced,0.11528063774108888
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,1,32,balanced,0.13776384353637697
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,1,32,balanced,0.050214400291442876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,1,32,balanced,0.1941823959350586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,1,32,balanced,0.05096320152282715
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,1,32,balanced,0.2350553512573242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,1,32,power_law_1.01,0.059921917915344236
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,1,32,power_law_1.01,0.05937920093536377
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,1,32,balanced,0.3403046417236328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,1,32,power_law_1.01,0.05972095966339112
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,1,32,balanced,0.4224063873291016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,1,32,power_law_1.01,0.08758527755737304
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,1,32,power_law_1.01,0.08911744117736817
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,1,32,power_law_1.01,0.09632639884948731
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,1,32,power_law_1.01,0.09738240242004395
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,1,32,power_law_1.01,0.09824383735656739
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,1,32,power_law_1.01,0.1040998363494873
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,1,32,power_law_1.01,0.10377728462219238
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,1,32,power_law_1.01,0.10769280433654785
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,1,32,power_law_1.01,0.11463808059692382
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,1,32,power_law_1.01,0.116245756149292
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,1,32,power_law_1.01,0.12032128334045411
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,1,32,power_law_1.01,0.13574015617370605
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,1,32,power_law_1.01,0.1289292812347412
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,1,32,power_law_1.01,0.18030080795288086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,1,32,power_law_1.01,0.20769023895263672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,1,32,power_law_1.01,0.2729023933410645
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,1,32,power_law_1.01,0.3537356948852539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,1,32,power_law_1.01,0.5567334365844727
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,4,8,power_law_1.2,0.5373772811889649
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,1,32,power_law_1.01,0.11467007637023925
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,1,32,power_law_1.01,0.768806381225586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,1,32,power_law_1.01,0.04117248058319092
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,4,8,power_law_1.2,0.8043545532226564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,1,32,power_law_1.01,1.010206756591797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,1,32,power_law_1.01,0.04129024028778076
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,1,32,power_law_1.01,0.041794562339782716
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,1,32,power_law_1.01,0.04775680065155029
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,1,32,power_law_1.01,0.04899199962615967
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,1,32,power_law_1.01,0.05059199810028077
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,1,32,power_law_1.01,0.0515993595123291
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,1,32,power_law_1.01,1.423171844482422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,1,32,power_law_1.01,0.052445440292358404
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,1,32,power_law_1.01,0.05412479877471924
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,1,32,power_law_1.01,0.054897918701171874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,1,32,power_law_1.01,0.058105602264404296
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,1,32,power_law_1.01,0.060751361846923826
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,1,32,power_law_1.01,0.06401792049407959
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,1,32,power_law_1.01,0.0647321605682373
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,1,32,power_law_1.01,0.062277121543884276
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,1,32,power_law_1.01,0.06620287895202637
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,1,32,power_law_1.01,0.06964992046356201
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,1,32,power_law_1.01,1.8651507568359373
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,1,32,power_law_1.01,0.09107968330383301
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,1,32,power_law_1.01,0.10636287689208986
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,1,32,power_law_1.01,0.1508236789703369
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,1,32,power_law_1.01,0.18708480834960936
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,1,32,power_law_1.01,0.276125431060791
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,1,32,power_law_1.01,0.3155353546142578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,1,32,power_law_1.2,0.06014976024627685
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,1,32,power_law_1.2,0.05958271980285644
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,1,32,power_law_1.01,0.5369139099121094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,1,32,power_law_1.01,2.9385061645507813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,1,32,power_law_1.2,0.08817024230957031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,1,32,power_law_1.2,0.08912511825561523
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,1,32,power_law_1.01,0.9285414123535156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,1,32,power_law_1.2,0.09814911842346191
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,1,32,power_law_1.2,0.09483263969421388
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,1,32,power_law_1.2,0.10262144088745118
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,1,32,power_law_1.2,0.10995712280273437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,1,32,power_law_1.2,0.11162879943847656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,1,32,power_law_1.01,1.373861083984375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,1,32,power_law_1.2,0.11515392303466795
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,1,32,power_law_1.2,0.12013695716857911
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,1,32,power_law_1.2,0.1228172779083252
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,1,32,power_law_1.2,0.05449728012084961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,1,32,power_law_1.2,0.1325708770751953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,1,32,power_law_1.01,0.6648908996582031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,4,8,power_law_1.2,0.7071564483642578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,1,32,power_law_1.2,0.1987571144104004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,1,32,power_law_1.2,0.287194881439209
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,1,32,power_law_1.2,0.10046079635620117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,1,32,power_law_1.2,0.3409292984008789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,1,32,power_law_1.2,0.5810636901855469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,1,32,power_law_1.2,0.4185497665405274
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,1,32,power_law_1.2,0.11676287651062012
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,1,32,power_law_1.2,0.04136320114135742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,1,32,power_law_1.2,0.0410752010345459
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,1,32,power_law_1.2,0.16053888320922852
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,1,32,power_law_1.2,0.039804160594940186
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,1,32,power_law_1.2,0.04779903888702393
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,1,32,power_law_1.2,0.04926464080810547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,1,32,power_law_1.2,0.8596428680419923
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,1,32,power_law_1.2,0.0508787202835083
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,1,32,power_law_1.2,0.05231359958648681
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,1,32,power_law_1.2,0.054448637962341305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,1,32,power_law_1.2,0.05534592151641846
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,1,32,power_law_1.2,0.05554687976837158
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,1,32,power_law_1.2,0.05924992084503174
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,1,32,power_law_1.2,0.06285183906555177
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,1,32,power_law_1.2,0.06402431964874268
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,1,32,power_law_1.2,0.06500864028930664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,1,32,power_law_1.2,0.06809088230133056
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,1,32,power_law_1.2,0.07218048095703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,1,32,power_law_1.2,0.08127488136291504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,1,32,power_law_1.2,0.10678527832031251
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,1,32,power_law_1.2,0.12072959899902344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,1,32,power_law_1.2,0.17008895874023439
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,1,32,power_law_1.2,0.2101862335205078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,1,32,power_law_1.2,0.3324620819091797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,1,32,power_law_1.2,1.273438720703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,1,32,power_law_1.2,0.44941951751708986
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,2,16,balanced,0.023596799373626708
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,2,16,balanced,0.02364032030105591
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,2,16,balanced,0.03069439888000488
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,2,16,balanced,0.04663551807403564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,2,16,balanced,0.05872640132904052
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,2,16,balanced,0.08929792404174805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,1,32,power_law_1.2,0.6264307022094726
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,2,16,balanced,0.09073663711547851
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,2,16,balanced,0.09330559730529785
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,2,16,balanced,0.09489151954650879
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,2,16,balanced,0.09707903861999512
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,2,16,balanced,0.09828479766845703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,2,16,balanced,0.1025715160369873
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,1,32,power_law_1.2,2.9071194458007814
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,2,16,balanced,0.10618623733520507
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,2,16,balanced,0.09968768119812012
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,2,16,balanced,0.10160639762878418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,2,16,balanced,0.10372351646423339
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,2,16,balanced,0.10959744453430176
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,2,16,balanced,0.11968000411987305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,2,16,balanced,0.12675711631774902
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,2,16,balanced,0.1573196792602539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,2,16,balanced,0.1658367919921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,2,16,balanced,0.23702016830444336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,1,32,power_law_1.2,1.8121612548828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,2,16,balanced,0.29537919998168943
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,2,16,balanced,0.41105022430419924
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,2,16,balanced,0.5182092666625977
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,2,16,balanced,0.025007359981536866
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,2,16,balanced,0.025296640396118165
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,2,16,balanced,0.02663424015045166
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,2,16,balanced,0.030133759975433348
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,2,16,balanced,0.9836774444580078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,2,16,balanced,0.04106880187988281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,2,16,balanced,0.051590399742126467
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,2,16,balanced,0.05342463970184326
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,2,16,balanced,0.053946881294250494
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,2,16,balanced,0.05621503829956055
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,2,16,balanced,0.05587200164794922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,2,16,balanced,0.058196477890014656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,2,16,balanced,0.06267007827758789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,2,16,balanced,0.7574259185791016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,2,16,balanced,0.06289792060852051
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,2,16,balanced,0.05805823802947998
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,2,16,balanced,0.06015615940093995
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,2,16,balanced,0.062255358695983885
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,2,16,balanced,0.06574336051940918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,2,16,balanced,0.0736844778060913
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,2,16,balanced,0.08040960311889649
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,2,16,balanced,0.10083711624145508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,2,16,balanced,0.11716223716735841
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,2,16,balanced,0.15512831687927248
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,2,16,balanced,0.18679168701171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,2,16,balanced,0.27406591415405274
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,1,32,power_law_1.2,0.758581771850586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,2,16,balanced,0.3408550262451172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,2,16,power_law_1.01,0.05737599849700927
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,2,16,balanced,0.499290885925293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,2,16,power_law_1.01,0.05698431968688965
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,2,16,power_law_1.01,0.05649919986724854
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,2,16,power_law_1.01,0.055324158668518066
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,2,16,balanced,0.6338150405883789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,2,16,power_law_1.01,0.0827660846710205
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,2,16,power_law_1.01,0.09107583999633789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,2,16,power_law_1.01,0.09403136253356933
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,2,16,power_law_1.01,0.09618816375732422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,2,16,power_law_1.01,0.09668992042541505
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,2,16,power_law_1.01,0.0994035243988037
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,2,16,power_law_1.01,0.10334207534790038
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,2,16,power_law_1.01,0.10608384132385254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,2,16,power_law_1.01,0.10156671524047851
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,2,16,power_law_1.01,0.10910847663879394
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,2,16,power_law_1.01,0.11203200340270995
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,2,16,power_law_1.01,0.12126848220825195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,2,16,power_law_1.01,0.14918272018432616
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,2,16,power_law_1.01,0.1625484848022461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,2,16,power_law_1.01,0.21185024261474608
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,2,16,power_law_1.01,0.2736947250366211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,2,16,power_law_1.01,0.08844032287597656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,2,16,power_law_1.01,0.3995904159545899
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,2,16,power_law_1.01,0.5084889602661133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,2,16,power_law_1.01,0.039883520603179935
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,2,16,power_law_1.01,0.770672607421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,2,16,power_law_1.01,0.04001023769378662
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,2,16,power_law_1.01,0.9615654754638673
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,2,16,power_law_1.01,0.03949055910110474
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,2,16,power_law_1.01,0.04086656093597412
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,2,16,power_law_1.01,0.04676864147186279
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,2,16,power_law_1.01,0.054096641540527346
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,2,16,power_law_1.01,0.05421696186065674
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,2,16,power_law_1.01,1.5684786987304686
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,2,16,power_law_1.01,0.05683072090148926
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,2,16,power_law_1.01,0.059357438087463375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,2,16,power_law_1.01,0.0593612813949585
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,2,16,power_law_1.01,0.06584320068359376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,2,16,power_law_1.01,0.061721601486206056
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,2,16,power_law_1.01,0.06033279895782471
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,2,16,power_law_1.01,0.06342144012451172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,2,16,power_law_1.01,0.06723840236663818
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,2,16,power_law_1.01,0.07268095970153808
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,2,16,power_law_1.01,0.08979968070983887
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,2,16,power_law_1.01,0.09843328475952148
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,2,16,power_law_1.01,0.1337715244293213
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,2,16,power_law_1.01,0.1705471992492676
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,2,16,power_law_1.01,0.21530368804931638
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,2,16,power_law_1.01,0.28771455764770504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,2,16,power_law_1.01,2.042170867919922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,2,16,power_law_1.01,0.5338560104370117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,2,16,power_law_1.01,0.05002880096435547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,2,16,power_law_1.2,0.05712895870208741
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,2,16,power_law_1.2,0.05472511768341064
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,2,16,power_law_1.01,0.9086950683593751
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,2,16,power_law_1.2,0.05012479782104492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,2,16,power_law_1.01,0.4200640106201172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,2,16,power_law_1.2,0.08340352058410645
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,2,16,power_law_1.01,1.1441792297363282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,2,16,power_law_1.2,0.08851840019226073
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,2,16,power_law_1.2,0.09165696144104005
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,2,16,power_law_1.2,0.09351807594299316
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,2,16,power_law_1.2,0.0961843204498291
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,2,16,power_law_1.2,0.09956095695495606
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,2,16,power_law_1.2,0.05340672016143799
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,2,16,power_law_1.2,0.10477312088012695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,2,16,power_law_1.2,0.10800895690917969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,2,16,power_law_1.2,0.10423551559448244
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,2,16,power_law_1.2,0.10812543869018554
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,2,16,power_law_1.2,0.1120627212524414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,2,16,power_law_1.2,0.11638143539428711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,2,16,power_law_1.2,0.1307430362701416
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,2,16,power_law_1.2,0.16105600357055666
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,2,16,power_law_1.2,0.1966054344177246
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,2,16,power_law_1.2,0.2631052780151367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,2,16,power_law_1.2,0.3373516845703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,2,16,power_law_1.2,0.49974014282226564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,2,16,power_law_1.2,0.6857561492919922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,2,16,power_law_1.2,0.03972863912582397
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,2,16,power_law_1.2,0.03798784017562866
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,2,16,power_law_1.2,0.03869951963424682
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,1,32,power_law_1.2,1.2505792236328124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,2,16,power_law_1.2,1.2308211517333985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,2,16,power_law_1.2,0.9995763397216797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,2,16,power_law_1.2,0.038755838871002195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,2,16,power_law_1.2,0.047411198616027835
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,2,16,power_law_1.2,0.05028351783752442
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,2,16,power_law_1.2,0.055617280006408684
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,2,16,power_law_1.2,0.05811327934265137
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,2,16,power_law_1.2,0.05860352039337158
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,2,16,power_law_1.2,0.06406655788421631
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,2,16,power_law_1.2,0.06788608074188232
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,2,16,power_law_1.2,0.06033152103424072
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,2,16,power_law_1.2,0.06444287776947022
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,2,16,power_law_1.2,0.06652031898498535
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,2,16,power_law_1.2,0.07041920185089111
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,2,16,power_law_1.2,0.07465343952178956
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,2,16,power_law_1.2,0.09783040046691895
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,2,16,power_law_1.2,0.11914752006530763
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,2,16,power_law_1.2,0.14813183784484862
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,2,16,power_law_1.2,0.052907519340515144
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,2,16,power_law_1.2,0.19193471908569334
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,2,16,power_law_1.2,0.28815359115600586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,2,16,power_law_1.2,0.3651174545288086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,2,16,power_law_1.2,0.46126976013183596
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,4,8,balanced,0.02244352102279663
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,4,8,balanced,0.023430399894714356
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,4,8,balanced,0.03627135992050171
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,4,8,balanced,0.04627200126647949
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,4,8,balanced,0.06146944046020508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,4,8,balanced,0.09050111770629883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,4,8,balanced,0.09547648429870606
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,4,8,balanced,0.09833087921142578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,4,8,balanced,0.10281984329223633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,4,8,balanced,0.1044940757751465
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,2,16,power_law_1.2,0.5819481658935547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,4,8,balanced,0.0968716812133789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,4,8,balanced,0.09920000076293946
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,4,8,balanced,0.10137855529785156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,4,8,balanced,0.10627200126647948
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,4,8,balanced,0.10925824165344238
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,4,8,balanced,0.11333888053894044
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,4,8,balanced,0.12190719604492188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,4,8,balanced,0.13797632217407227
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,4,8,balanced,0.15292415618896485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,4,8,balanced,0.19012992858886718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,4,8,balanced,0.21427839279174804
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,4,8,balanced,0.3098969650268555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,2,16,power_law_1.2,1.915820770263672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,4,8,balanced,0.396234245300293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,4,8,balanced,0.5580467224121094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,4,8,balanced,0.7147328186035156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,4,8,balanced,0.02439039945602417
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,4,8,balanced,1.0520806121826172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,4,8,balanced,0.02500607967376709
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,4,8,balanced,0.025743360519409182
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,4,8,balanced,0.030168321132659916
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,4,8,balanced,0.04473087787628174
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,4,8,balanced,0.05813504219055175
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,4,8,balanced,0.05896063804626465
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,4,8,balanced,0.06110208034515381
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,4,8,balanced,1.3671347045898439
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,4,8,balanced,0.06425600051879883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,4,8,balanced,0.06593791961669922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,4,8,balanced,0.05999743938446045
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,4,8,balanced,0.06170368194580078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,4,8,balanced,0.06322432041168213
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,4,8,balanced,0.06679168224334717
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,4,8,balanced,0.06858367919921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,4,8,balanced,0.07184895992279053
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,4,8,balanced,0.08004608154296874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,4,8,balanced,0.09683327674865723
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,4,8,balanced,0.1119654369354248
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,4,8,balanced,0.1412070369720459
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,4,8,balanced,0.17018623352050782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,4,8,balanced,0.23551231384277344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,4,8,balanced,0.2912704086303711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,4,8,balanced,0.43533184051513674
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,4,8,balanced,0.5550540924072266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,4,8,balanced,0.8194764709472656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,4,8,power_law_1.01,0.045320959091186525
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,4,8,balanced,1.0573836517333983
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,4,8,power_law_1.01,0.04533760070800781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,1,32,power_law_1.2,4.112459411621094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,4,8,power_law_1.01,0.050402560234069825
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,4,8,power_law_1.01,0.05611135959625244
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,4,8,power_law_1.01,0.05569920063018798
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,4,8,power_law_1.01,0.0857203197479248
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,4,8,power_law_1.01,0.09008255958557129
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,4,8,power_law_1.01,0.09475839614868165
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,4,8,power_law_1.01,0.09786239624023438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,4,8,power_law_1.01,0.09950847625732422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,4,8,power_law_1.01,0.10392064094543456
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,4,8,power_law_1.01,0.10338047981262206
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,4,8,power_law_1.01,0.101461763381958
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,4,8,power_law_1.01,0.10534015655517578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,4,8,power_law_1.01,0.11027711868286132
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,4,8,power_law_1.01,0.11373696327209473
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,4,8,power_law_1.01,0.12311936378479005
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,1,32,power_law_1.2,1.844523468017578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,4,8,power_law_1.01,0.1507148838043213
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,4,8,power_law_1.01,0.2187123107910156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,4,8,power_law_1.01,0.2516057586669922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,4,8,power_law_1.01,0.36323200225830077
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,4,8,power_law_1.01,0.4941862487792969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,4,8,power_law_1.01,0.7106265258789063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,4,8,power_law_1.01,0.1334284782409668
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,4,8,power_law_1.01,0.02743168115615845
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,4,8,power_law_1.01,0.9263142395019532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,4,8,power_law_1.01,0.03658495903015137
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,4,8,power_law_1.01,0.04206975936889649
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,4,8,power_law_1.01,0.041864957809448246
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,4,8,power_law_1.01,1.4072933959960938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,4,8,power_law_1.01,0.05497471809387207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,4,8,power_law_1.01,0.057090559005737306
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,4,8,power_law_1.01,0.05893887996673584
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,4,8,power_law_1.01,0.061735677719116214
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,4,8,power_law_1.01,0.06435200214385986
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,4,8,power_law_1.01,0.06803711891174316
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,4,8,power_law_1.01,1.8232371520996093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,4,8,power_law_1.01,0.06303872108459473
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,4,8,power_law_1.01,0.03027967929840088
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,4,8,power_law_1.01,0.06449664115905762
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,4,8,power_law_1.01,0.06698495864868163
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,4,8,power_law_1.01,0.06944128036499023
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,4,8,power_law_1.01,0.07773695945739746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,4,8,power_law_1.01,0.09175807952880859
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,2,16,power_law_1.2,1.053222427368164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,4,8,power_law_1.01,0.10690943717956543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,4,8,power_law_1.01,0.13522175788879393
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,4,8,power_law_1.01,0.1635647964477539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,4,8,power_law_1.01,0.25322751998901366
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,4,8,power_law_1.01,0.3250931167602539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,4,8,power_law_1.01,0.42258430480957027
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,4,8,power_law_1.01,0.06137343883514405
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,4,8,power_law_1.2,0.04518784046173095
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,4,8,power_law_1.01,0.6038067245483398
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,4,8,power_law_1.2,0.04572415828704834
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,4,8,power_law_1.2,0.04882815837860107
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,4,8,power_law_1.2,0.05359488010406495
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,4,8,power_law_1.01,0.837386245727539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,4,8,power_law_1.2,0.08668031692504882
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,4,8,power_law_1.2,0.05459712028503418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,4,8,power_law_1.2,0.09513216018676758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,4,8,power_law_1.2,0.10136447906494142
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,4,8,power_law_1.2,0.10582143783569335
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,4,8,power_law_1.2,0.10274304389953612
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,4,8,power_law_1.2,0.10268799781799316
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,4,8,power_law_1.2,0.11257984161376952
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,4,8,power_law_1.2,0.11583744049072267
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,4,8,power_law_1.2,0.12313856124877928
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,4,8,power_law_1.2,0.13966464042663573
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,4,8,power_law_1.2,0.16298240661621094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,4,8,power_law_1.2,0.09148799896240234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,4,8,power_law_1.2,0.2330252838134766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,4,8,power_law_1.2,0.09960960388183594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,4,8,power_law_1.01,1.0512895965576172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,4,8,power_law_1.2,0.2847001647949219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,4,8,power_law_1.2,0.10705023765563966
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,4,8,power_law_1.2,0.407586555480957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,4,8,power_law_1.2,0.5528422546386719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,4,8,power_law_1.2,0.027294719219207765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,4,8,power_law_1.2,0.02796799898147583
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,4,8,power_law_1.2,0.03540735960006714
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,4,8,power_law_1.2,0.04115583896636963
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,4,8,power_law_1.2,0.03852799892425537
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,4,8,power_law_1.2,0.7418342590332031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,4,8,power_law_1.2,0.05534207820892334
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,4,8,power_law_1.2,0.05739776134490967
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,4,8,power_law_1.2,0.05920767784118652
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,4,8,power_law_1.2,0.06275328159332275
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,4,8,power_law_1.2,0.06419328212738037
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,4,8,power_law_1.2,0.06969088077545166
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,4,8,power_law_1.2,0.0615065622329712
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,4,8,power_law_1.2,0.06599552154541015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,4,8,power_law_1.2,0.06515583992004395
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,4,8,power_law_1.2,0.06687744140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,4,8,power_law_1.2,0.07118207931518554
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,4,8,power_law_1.2,1.3883187866210938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,4,8,power_law_1.2,0.07690239906311035
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,4,8,power_law_1.2,0.09889535903930664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,4,8,power_law_1.2,0.11047167778015136
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,4,8,power_law_1.2,0.14926464080810548
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,4,8,power_law_1.2,0.18526208877563477
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,4,8,power_law_1.2,0.2636748886108399
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,4,8,power_law_1.2,0.34060672760009764
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,4,8,power_law_1.2,0.49260032653808594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,8,4,balanced,0.02349951982498169
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,8,4,balanced,0.02509567975997925
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,8,4,balanced,0.03807487964630127
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,8,4,balanced,0.04801792144775391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,8,4,balanced,0.06295680046081543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,8,4,balanced,0.0664409589767456
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,8,4,balanced,0.0955737590789795
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,8,4,balanced,0.09877247810363769
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,8,4,balanced,0.10064512252807618
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,8,4,balanced,0.10383359909057617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,8,4,balanced,0.09574527740478515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,8,4,balanced,0.09818880081176758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,8,4,balanced,0.10048000335693359
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,8,4,balanced,0.10286975860595704
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,8,4,balanced,0.10647295951843261
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,8,4,balanced,0.10981632232666017
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,8,4,balanced,0.11783295631408693
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,2,16,power_law_1.2,2.761797180175781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,8,4,balanced,0.13272319793701173
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,8,4,balanced,0.1473369598388672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,8,4,balanced,0.18134143829345703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,8,4,balanced,0.20871936798095705
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,8,4,balanced,0.3003647994995117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,4,8,power_law_1.2,0.9816268920898438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,8,4,balanced,0.38467201232910153
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,8,4,balanced,0.550519027709961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,8,4,balanced,0.7100633239746094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,8,4,balanced,0.025201919078826907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,8,4,balanced,1.0449612426757813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,8,4,balanced,0.025566720962524415
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,8,4,balanced,0.028090879917144772
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,2,16,power_law_1.2,1.3681765747070311
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,8,4,balanced,0.051582717895507814
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,8,4,balanced,0.05594240188598633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,8,4,balanced,0.06874623775482178
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,8,4,balanced,1.3788275146484374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,8,4,balanced,0.07196800231933594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,8,4,balanced,0.07307263851165771
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,8,4,balanced,0.07593728065490722
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,8,4,balanced,0.06784639835357667
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,8,4,balanced,0.07043712139129639
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,8,4,balanced,0.0715174388885498
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,8,4,balanced,0.07474944114685059
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,8,4,balanced,0.07614079952239991
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,8,4,balanced,0.0793062400817871
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,8,4,balanced,0.03523328065872192
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,8,4,balanced,0.08539648056030273
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,8,4,balanced,0.10095999717712403
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,8,4,balanced,0.11437696456909179
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,8,4,balanced,0.14288895606994628
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,8,4,balanced,0.16777215957641603
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,8,4,balanced,0.2407206344604492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,8,4,balanced,0.2933606338500977
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,8,4,balanced,0.43698558807373045
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,8,4,balanced,0.5597824096679688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,8,4,power_law_1.01,0.03061887979507446
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,8,4,power_law_1.01,0.043359999656677244
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,4,8,power_law_1.2,0.6212774276733398
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,8,4,power_law_1.01,0.045528321266174315
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,8,4,balanced,0.8252134704589844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,8,4,power_law_1.01,0.04967296123504639
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,8,4,power_law_1.01,0.05422207832336425
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,8,4,power_law_1.01,0.061607680320739745
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,8,4,balanced,1.0923161315917969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,8,4,power_law_1.01,0.09656319618225098
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,8,4,power_law_1.01,0.09900671958923339
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,8,4,power_law_1.01,0.10152704238891601
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,8,4,power_law_1.01,0.10124799728393555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,8,4,power_law_1.01,0.09757568359374999
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,8,4,power_law_1.01,0.09984255790710449
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,8,4,power_law_1.01,0.10291968345642091
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,8,4,power_law_1.01,0.10665727615356446
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,8,4,power_law_1.01,0.11192319869995118
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,8,4,power_law_1.01,0.11901439666748045
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,8,4,power_law_1.01,0.1348915195465088
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,8,4,power_law_1.01,0.1509823989868164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,8,4,power_law_1.01,0.18903295516967772
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,8,4,power_law_1.01,0.2472652816772461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,8,4,power_law_1.01,0.09268351554870605
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,8,4,power_law_1.01,0.3429465484619141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,8,4,power_law_1.01,0.4114060974121093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,8,4,power_law_1.01,0.026725120544433593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,8,4,power_law_1.01,0.6278911972045899
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,8,4,power_law_1.01,0.03330816030502319
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,8,4,power_law_1.01,0.8021734619140626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,8,4,power_law_1.01,0.0348799991607666
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,8,4,power_law_1.01,0.04025343894958496
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,8,4,power_law_1.01,0.04005760192871094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,8,4,power_law_1.01,1.225418243408203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,8,4,power_law_1.01,0.06882431983947754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,8,4,power_law_1.01,0.07162496089935302
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,8,4,power_law_1.01,1.6124403381347654
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,8,4,power_law_1.01,0.0748799991607666
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,8,4,power_law_1.01,0.0753766393661499
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,8,4,power_law_1.01,0.06782720088958741
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,8,4,power_law_1.01,0.07040639877319335
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,8,4,power_law_1.01,0.07383679866790771
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,8,4,power_law_1.01,0.0656934404373169
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,8,4,power_law_1.01,0.0753062391281128
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,8,4,power_law_1.01,0.07883903980255128
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,8,4,power_law_1.01,0.0871731185913086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,8,4,power_law_1.01,0.10294655799865722
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,8,4,power_law_1.01,0.1220736026763916
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,8,4,power_law_1.01,0.15272192001342771
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,8,4,power_law_1.01,0.18776063919067382
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,8,4,power_law_1.01,0.26812927246093754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,8,4,power_law_1.01,0.3323276901245117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,8,4,power_law_1.01,0.4901132965087891
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,8,4,power_law_1.01,0.6329612731933594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,8,4,power_law_1.2,0.025671679973602295
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,8,4,power_law_1.01,0.9093824005126953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,8,4,power_law_1.2,0.04067840099334717
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,8,4,power_law_1.2,0.049800958633422855
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,8,4,power_law_1.2,0.04655104160308838
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,8,4,power_law_1.01,1.1873216247558593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,8,4,power_law_1.2,0.062391037940979
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,8,4,power_law_1.2,0.09310336112976074
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,8,4,power_law_1.2,0.09642751693725586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,8,4,power_law_1.2,0.10025216102600099
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,8,4,power_law_1.2,0.10253312110900878
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,8,4,power_law_1.2,0.10033791542053223
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,8,4,power_law_1.2,0.10108544349670409
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,8,4,power_law_1.2,0.04313983917236328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,8,4,power_law_1.2,0.10434432029724121
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,8,4,power_law_1.2,0.10749695777893067
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,8,4,power_law_1.2,0.1124608039855957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,8,4,power_law_1.2,0.11995648384094237
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,8,4,power_law_1.2,0.13766400337219237
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,8,4,power_law_1.01,0.05411456108093262
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,8,4,power_law_1.2,0.15652607917785644
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,8,4,power_law_1.2,0.19424383163452147
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,8,4,power_law_1.2,0.2551219177246094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,8,4,power_law_1.2,0.09852928161621094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,8,4,power_law_1.2,0.3475187301635742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,8,4,power_law_1.2,0.4797760009765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,8,4,power_law_1.2,0.026758399009704593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,8,4,power_law_1.2,0.03222143888473511
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,4,8,power_law_1.2,0.9238579559326172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,8,4,power_law_1.2,0.035278079509735105
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,8,4,power_law_1.2,0.03868671894073487
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,8,4,power_law_1.2,0.6722278594970703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,8,4,power_law_1.2,0.04519936084747315
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,8,4,power_law_1.2,0.06663551807403564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,8,4,power_law_1.2,0.06969600200653077
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,8,4,power_law_1.2,0.07230463981628418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,8,4,power_law_1.2,0.07438975811004639
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,8,4,power_law_1.2,0.07138175964355468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,8,4,power_law_1.2,0.06933760166168212
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,8,4,power_law_1.2,0.07032959938049317
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,8,4,power_law_1.2,0.074170880317688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,8,4,power_law_1.2,0.07617663860321044
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,8,4,power_law_1.2,0.08018688201904298
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,8,4,power_law_1.2,0.0868671989440918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,8,4,power_law_1.2,0.1051699161529541
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,8,4,power_law_1.2,0.12226816177368165
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,8,4,power_law_1.2,0.15619839668273924
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,8,4,power_law_1.2,0.8324454498291015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,8,4,power_law_1.2,0.20055423736572267
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,8,4,power_law_1.2,0.2728819274902344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,8,4,power_law_1.2,0.3503936004638672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,8,4,power_law_1.2,0.05268608093261719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,4,8,power_law_1.2,1.182992630004883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,8,4,power_law_1.2,0.496808967590332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,16,2,balanced,0.029834239482879638
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,16,2,balanced,0.039244799613952636
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,16,2,balanced,0.02338304042816162
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,4,8,power_law_1.2,2.0118118286132813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,16,2,balanced,0.05099647998809814
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,16,2,balanced,0.06828288078308106
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,16,2,balanced,0.07180543899536132
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,16,2,balanced,0.07276159763336182
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,16,2,balanced,0.09567872047424317
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,16,2,balanced,0.09616895675659179
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,16,2,balanced,0.09747200012207032
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,16,2,balanced,0.09965439796447753
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,16,2,balanced,0.10170880317687989
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,16,2,balanced,0.10478848457336425
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,16,2,balanced,0.1076364803314209
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,16,2,balanced,0.11063039779663086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,16,2,balanced,0.11755392074584961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,16,2,balanced,0.13110527992248536
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,16,2,balanced,0.0736243200302124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,16,2,balanced,0.14553215980529785
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,16,2,balanced,0.18411136627197267
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,16,2,balanced,0.21052799224853516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,16,2,balanced,0.2829094314575195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,16,2,balanced,0.38742145538330075
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,16,2,balanced,0.5563865661621094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,16,2,balanced,0.7394956970214844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,16,2,balanced,0.024358398914337158
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,16,2,balanced,1.0669273376464843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,16,2,balanced,0.02615808010101318
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,16,2,balanced,0.02984832048416138
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,16,2,balanced,0.04010496139526367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,16,2,balanced,0.06069632053375244
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,16,2,balanced,0.061835517883300786
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,8,4,power_law_1.2,1.261538543701172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,16,2,balanced,0.06280447959899901
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,16,2,balanced,0.06363647937774658
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,16,2,balanced,0.08014719963073731
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,16,2,balanced,0.08128255844116211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,16,2,balanced,0.08286720275878906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,16,2,balanced,1.4327539062499999
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,16,2,balanced,0.08531455993652344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,16,2,balanced,0.08581376075744629
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,16,2,balanced,0.09099519729614258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,16,2,balanced,0.08973055839538574
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,16,2,balanced,0.09337087631225585
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,16,2,balanced,0.09836031913757323
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,16,2,balanced,0.11097087860107421
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,16,2,balanced,0.12240127563476562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,16,2,balanced,0.14995583534240722
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,16,2,balanced,0.17507455825805665
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,16,2,balanced,0.2552921676635742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,16,2,balanced,0.3030873680114746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,16,2,balanced,0.4545139312744141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,16,2,balanced,0.5838399887084961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,16,2,power_law_1.01,0.024294400215148927
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,16,2,power_law_1.01,0.039701759815216064
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,16,2,power_law_1.01,0.04455552101135254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,16,2,power_law_1.01,0.04991615772247314
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,16,2,power_law_1.01,0.05534592151641846
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,16,2,power_law_1.01,0.0653004789352417
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,16,2,balanced,1.1516774749755858
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,16,2,power_law_1.01,0.06983935832977295
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,16,2,power_law_1.01,0.09347968101501465
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,16,2,power_law_1.01,0.09466496467590332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,16,2,power_law_1.01,0.09641983985900879
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,16,2,balanced,0.8680614471435547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,16,2,power_law_1.01,0.09919615745544433
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,16,2,power_law_1.01,0.10110207557678223
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,16,2,power_law_1.01,0.10852224349975585
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,16,2,power_law_1.01,0.11192959785461425
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,16,2,power_law_1.01,0.06798848152160644
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,16,2,power_law_1.01,0.1192191982269287
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,16,2,power_law_1.01,0.13639936447143555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,8,4,power_law_1.2,0.6317043304443359
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,16,2,power_law_1.01,0.1888025665283203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,16,2,power_law_1.01,0.22455680847167966
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,16,2,power_law_1.01,0.3384601593017578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,16,2,power_law_1.01,0.4280524826049804
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,16,2,power_law_1.01,0.10486144065856932
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,16,2,power_law_1.01,0.605593605041504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,16,2,power_law_1.01,0.02550015926361084
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,16,2,power_law_1.01,0.15478143692016602
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,16,2,power_law_1.01,0.7703539276123047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,16,2,power_law_1.01,0.028876800537109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,16,2,power_law_1.01,0.03591808080673218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,16,2,power_law_1.01,0.04023039817810058
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,16,2,power_law_1.01,0.04637311935424805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,16,2,power_law_1.01,0.06256000041961671
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,16,2,power_law_1.01,1.1570022583007813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,16,2,power_law_1.01,0.05948287963867187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,16,2,power_law_1.01,0.06115583896636963
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,16,2,power_law_1.01,0.0765555191040039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,16,2,power_law_1.01,0.078919677734375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,16,2,power_law_1.01,0.0803276824951172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,16,2,power_law_1.01,1.4791693115234374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,16,2,power_law_1.01,0.0817369556427002
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,16,2,power_law_1.01,0.08427904129028321
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,16,2,power_law_1.01,0.08826751708984375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,16,2,power_law_1.01,0.08986495971679688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,16,2,power_law_1.01,0.10158975601196289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,16,2,power_law_1.01,0.11803520202636719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,16,2,power_law_1.01,0.13599743843078613
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,16,2,power_law_1.01,0.16929536819458008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,16,2,power_law_1.01,0.2043059158325195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,16,2,power_law_1.01,0.295281925201416
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,16,2,power_law_1.01,0.3619404983520508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,16,2,power_law_1.01,0.5089344024658203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,16,2,power_law_1.01,0.6484172821044922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,16,2,power_law_1.01,0.09383935928344726
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,16,2,power_law_1.2,0.03944960117340088
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,16,2,power_law_1.2,0.04354047775268555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,16,2,power_law_1.01,0.9523763275146484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,16,2,power_law_1.2,0.04849535942077637
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,16,2,power_law_1.2,0.050402560234069825
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,16,2,power_law_1.2,0.06862976074218749
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,16,2,power_law_1.2,0.06776319980621338
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,16,2,power_law_1.2,0.07184895992279053
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,16,2,power_law_1.01,1.2371443176269532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,16,2,power_law_1.2,0.09413120269775391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,16,2,power_law_1.2,0.09473535537719727
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,16,2,power_law_1.2,0.0963161563873291
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,16,2,power_law_1.2,0.0988428783416748
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,16,2,power_law_1.2,0.10648320198059083
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,16,2,power_law_1.2,0.10950783729553222
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,16,2,power_law_1.2,0.11263104438781739
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,16,2,power_law_1.2,0.12086912155151368
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,16,2,power_law_1.2,0.13906559944152833
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,16,2,power_law_1.2,0.15714943885803223
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,16,2,power_law_1.2,0.19080320358276368
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,16,2,power_law_1.2,0.2251737594604492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,16,2,power_law_1.2,0.10121983528137206
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,16,2,power_law_1.2,0.33619583129882813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,16,2,power_law_1.2,0.02443135976791382
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,16,2,power_law_1.2,0.43976318359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,8,4,power_law_1.2,0.9427417755126953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,16,2,power_law_1.2,0.025379838943481448
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,16,2,power_law_1.2,0.6223424148559571
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,16,2,power_law_1.2,0.031303679943084715
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,16,2,power_law_1.2,0.03508352041244507
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,16,2,power_law_1.2,0.04086527824401855
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,16,2,power_law_1.2,0.061066241264343256
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,16,2,power_law_1.2,0.06058495998382568
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,16,2,power_law_1.2,0.06127615928649902
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,16,2,power_law_1.2,0.0770905590057373
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,16,2,power_law_1.2,0.07877503871917725
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,16,2,power_law_1.2,0.07948416233062744
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,16,2,power_law_1.2,0.08368767738342285
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,16,2,power_law_1.2,0.039647998809814455
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,16,2,power_law_1.2,0.08887680053710938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,16,2,power_law_1.2,0.0909273624420166
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,16,2,power_law_1.2,0.0949068832397461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,16,2,power_law_1.2,0.10226048469543456
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,16,2,power_law_1.2,0.12088319778442383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,16,2,power_law_1.2,0.13690752029418946
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,16,2,power_law_1.2,0.17173248291015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,16,2,power_law_1.2,0.7719641876220703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,16,2,power_law_1.2,0.20601472854614258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,16,2,power_law_1.2,0.2984486389160156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,16,2,power_law_1.2,0.08493184089660645
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,16,2,power_law_1.2,0.37336704254150394
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,8,4,power_law_1.2,1.2021171569824218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,32,1,balanced,0.030443520545959474
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,32,1,balanced,0.03747519969940186
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,32,1,balanced,0.052596478462219244
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,32,1,balanced,0.06575232028961182
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,32,1,balanced,0.09036416053771973
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,32,1,balanced,0.09192319869995116
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,32,1,balanced,0.09354432106018067
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,32,1,balanced,0.09493375778198242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,16,2,power_law_1.2,0.5106227111816406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,32,1,balanced,0.09693375587463379
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,32,1,balanced,0.0976416015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,32,1,balanced,0.10080256462097167
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,8,4,power_law_1.2,1.6862413024902345
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,32,1,balanced,0.11871359825134278
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,32,1,balanced,0.11980928421020506
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,32,1,balanced,0.12429632186889647
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,32,1,balanced,0.12626688003540038
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,32,1,balanced,0.13162240028381347
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,32,1,balanced,0.143503360748291
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,32,1,balanced,0.15652928352355958
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,32,1,balanced,0.20039424896240235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,32,1,balanced,0.22408384323120117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,32,1,balanced,0.31654144287109376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,32,1,balanced,0.12327487945556641
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,32,1,balanced,0.4111616134643555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,32,1,balanced,0.6083667373657227
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,32,1,balanced,0.8009375762939452
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,32,1,balanced,0.03221888065338135
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,16,2,power_law_1.2,1.190772476196289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,32,1,balanced,0.03630847930908203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,32,1,balanced,1.2001036834716796
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,32,1,balanced,0.06199808120727539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,32,1,balanced,0.09626367568969726
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,32,1,balanced,0.09890432357788086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,32,1,balanced,0.098274564743042
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,32,1,balanced,0.09938559532165528
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,32,1,balanced,0.10009728431701662
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,32,1,balanced,0.10045439720153808
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,32,1,balanced,0.10224127769470215
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,32,1,balanced,0.04434559822082519
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,32,1,balanced,0.12740415573120117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,32,1,balanced,0.12791168212890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,32,1,balanced,0.12997952461242676
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,32,1,balanced,0.13129728317260742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,32,1,balanced,0.1323744010925293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,32,1,balanced,0.1376319980621338
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,32,1,balanced,0.1493561553955078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,32,1,balanced,0.16120832443237304
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,32,1,balanced,0.184913272857666
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,32,1,balanced,0.21030527114868164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,32,1,balanced,1.6267379760742187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,32,1,balanced,0.33670207977294925
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,16,2,power_law_1.2,0.6596543884277344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,32,1,balanced,0.3812076950073242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,32,1,balanced,0.5584070587158203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,32,1,power_law_1.01,0.030152320861816406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,32,1,balanced,0.73614013671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,32,1,power_law_1.01,0.04609024047851563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,32,1,power_law_1.01,0.052768640518188484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,32,1,power_law_1.01,0.060232319831848145
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,32,1,power_law_1.01,0.08451264381408692
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,32,1,balanced,1.082877426147461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,32,1,power_law_1.01,0.08898303985595704
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,32,1,power_law_1.01,0.09112447738647461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,32,1,power_law_1.01,0.09503040313720704
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,32,1,balanced,1.4428997802734376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,32,1,power_law_1.01,0.09739775657653808
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,32,1,power_law_1.01,0.10408448219299316
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,32,1,power_law_1.01,0.11786368370056152
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,32,1,power_law_1.01,0.11960831642150879
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,32,1,power_law_1.01,0.1284012794494629
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,32,1,power_law_1.01,0.07119679927825928
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,32,1,power_law_1.01,0.13111871719360352
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,32,1,power_law_1.01,0.139237117767334
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,32,1,power_law_1.01,0.15544511795043944
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,32,1,power_law_1.01,0.172740478515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,32,1,power_law_1.01,0.2093440055847168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,32,1,power_law_1.01,0.24924224853515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,32,1,power_law_1.01,0.3482828903198242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,32,1,power_law_1.01,0.46593280792236325
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,32,1,power_law_1.01,0.653016357421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,32,1,power_law_1.01,0.12337471961975097
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,32,1,power_law_1.01,0.8393644714355469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,32,1,power_law_1.01,0.03223743915557861
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,32,1,power_law_1.01,0.03653503894805908
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,32,1,power_law_1.01,0.0477177619934082
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,32,1,power_law_1.01,0.056029438972473145
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,32,1,power_law_1.01,0.06755199909210205
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,32,1,power_law_1.01,0.0900812816619873
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,32,1,power_law_1.01,1.2377823638916017
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,32,1,power_law_1.01,0.09251903533935547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,32,1,power_law_1.01,0.09509823799133302
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,32,1,power_law_1.01,0.09552895545959472
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,32,1,power_law_1.01,0.09915519714355468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,32,1,power_law_1.01,0.12345215797424317
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,32,1,power_law_1.01,0.12514944076538087
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,32,1,power_law_1.01,1.6456838989257812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,32,1,power_law_1.01,0.1311392021179199
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,32,1,power_law_1.01,0.133755521774292
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,32,1,power_law_1.01,0.1402233600616455
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,32,1,power_law_1.01,0.08629119873046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,32,1,power_law_1.01,0.1597433567047119
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,32,1,power_law_1.01,0.18324607849121094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,32,1,power_law_1.01,0.20221567153930664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,32,1,power_law_1.01,0.23956480026245117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,32,1,power_law_1.01,0.28293760299682613
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,32,1,power_law_1.01,0.38022014617919925
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,32,1,power_law_1.01,0.45428287506103515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,32,1,power_law_1.01,0.62787841796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,32,1,power_law_1.2,0.03048320055007935
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,32,1,power_law_1.01,0.803881607055664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,32,1,power_law_1.2,0.04605696201324463
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,32,1,power_law_1.2,0.05255296230316162
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,32,1,power_law_1.2,0.058039679527282714
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,32,1,power_law_1.2,0.06845823764801026
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,32,1,power_law_1.2,0.0886451244354248
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,32,1,power_law_1.01,1.1613587188720704
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,32,1,power_law_1.2,0.08895872116088868
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,32,1,power_law_1.2,0.09229248046875001
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,32,1,power_law_1.2,0.09637568473815919
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,32,1,power_law_1.2,0.09829952239990233
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,32,1,power_law_1.2,0.10425663948059083
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,32,1,power_law_1.2,0.11828672409057617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,32,1,power_law_1.01,1.5198733520507812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,32,1,power_law_1.2,0.12596991539001465
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,32,1,power_law_1.2,0.1283852767944336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,32,1,power_law_1.2,0.13228032112121582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,32,1,power_law_1.2,0.1404620838165283
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,32,1,power_law_1.2,0.15672063827514648
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,32,1,power_law_1.2,0.17431808471679688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,32,1,power_law_1.2,0.21128639221191406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,16,2,power_law_1.2,0.9587789154052734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,32,1,power_law_1.2,0.24938623428344728
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,32,1,power_law_1.2,0.3536934280395508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,32,1,power_law_1.2,0.11989567756652833
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,32,1,power_law_1.2,0.03205183982849121
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,32,1,power_law_1.2,0.03707456111907959
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,32,1,power_law_1.2,0.04650239944458008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,32,1,power_law_1.2,0.05525311946868896
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,32,1,power_law_1.2,0.06482560157775878
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,32,1,power_law_1.2,0.09044416427612305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,32,1,power_law_1.2,0.09083968162536621
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,32,1,power_law_1.2,0.09375679969787598
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,32,1,power_law_1.2,0.4757004928588867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,16,2,power_law_1.2,1.2466841888427733
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,32,1,power_law_1.2,0.09606464385986328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,32,1,power_law_1.2,0.0961228847503662
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,32,1,power_law_1.2,0.09901056289672852
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,32,1,power_law_1.2,0.1256588840484619
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,32,1,power_law_1.2,0.12712063789367675
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,32,1,power_law_1.2,0.6564569854736328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,32,1,power_law_1.2,0.13318143844604494
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,32,1,power_law_1.2,0.1398911952972412
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,32,1,power_law_1.2,0.14696384429931642
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,32,1,power_law_1.2,0.16770431518554688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,32,1,power_law_1.2,0.18552448272705077
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,32,1,power_law_1.2,0.20597440719604493
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,32,1,power_law_1.2,0.2466099166870117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,16,2,power_law_1.2,1.5028775024414063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,32,1,power_law_1.2,0.2839033508300781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,32,1,power_law_1.2,0.3808403015136719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,32,1,power_law_1.2,0.45950782775878907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,1,32,balanced,0.06716544151306152
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,1,32,balanced,0.06684544086456298
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,1,32,balanced,0.06694655895233155
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,1,32,balanced,0.07905151844024658
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,32,1,power_law_1.2,0.6322745513916016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,1,32,balanced,0.1090329647064209
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,1,32,balanced,0.18445056915283203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,1,32,balanced,0.1855449676513672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,1,32,balanced,0.1891391944885254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,1,32,balanced,0.18591487884521485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,1,32,balanced,0.19503744125366212
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,1,32,balanced,0.18696704864501953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,1,32,balanced,0.19077888488769532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,1,32,balanced,0.19680255889892578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,1,32,balanced,0.20178688049316404
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,32,1,power_law_1.2,0.8383097839355468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,1,32,balanced,0.20031999588012694
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,1,32,balanced,0.2092812728881836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,1,32,balanced,0.20802560806274414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,1,32,balanced,0.2300556755065918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,1,32,balanced,0.2464614486694336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,1,32,balanced,0.3133465576171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,1,32,balanced,0.3477056121826172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,1,32,balanced,0.49820671081542967
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,1,32,balanced,0.57185791015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,1,32,balanced,0.7288217926025391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,1,32,balanced,0.9612556457519531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,1,32,balanced,0.040687360763549806
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,1,32,balanced,0.04071807861328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,1,32,balanced,1.377274932861328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,1,32,balanced,0.04053376197814941
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,1,32,balanced,0.04943615913391113
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,1,32,balanced,0.0688972806930542
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,1,32,balanced,0.08451328277587891
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,1,32,balanced,1.7948045349121096
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,1,32,balanced,0.08860416412353515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,1,32,balanced,0.08812543869018555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,1,32,balanced,0.08819328308105469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,1,32,balanced,0.09150976181030274
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,1,32,balanced,0.09364864349365234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,32,1,power_law_1.2,0.8102950286865234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,1,32,balanced,0.09625344276428223
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,1,32,balanced,0.10094719886779786
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,1,32,balanced,0.10429696083068847
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,1,32,balanced,0.10571392059326172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,1,32,balanced,0.11493247985839843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,1,32,balanced,0.12359680175781249
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,1,32,balanced,0.1410636806488037
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,1,32,balanced,0.15097984313964843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,1,32,balanced,0.21180671691894531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,1,32,balanced,0.2410176086425781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,1,32,balanced,0.10899200439453124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,1,32,balanced,0.33590145111083985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,1,32,balanced,0.4297792053222656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,1,32,power_law_1.01,0.12843008041381837
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,1,32,balanced,0.5595135879516602
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,1,32,balanced,0.7622029113769531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,1,32,power_law_1.01,0.12806015968322754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,1,32,power_law_1.01,0.12450559616088867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,1,32,power_law_1.01,0.1563302421569824
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,1,32,power_law_1.01,0.17091455459594726
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,1,32,power_law_1.01,0.17364864349365233
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,1,32,power_law_1.01,0.1739072036743164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,1,32,power_law_1.01,0.18368000030517578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,1,32,power_law_1.01,0.18958463668823242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,1,32,balanced,0.09331839561462403
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,1,32,power_law_1.01,0.19136512756347657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,1,32,power_law_1.01,0.20045568466186525
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,1,32,power_law_1.01,0.20692863464355468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,1,32,power_law_1.01,0.22085376739501955
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,1,32,power_law_1.01,0.24347391128540039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,1,32,power_law_1.01,0.25694080352783205
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,1,32,power_law_1.01,0.2924812889099121
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,1,32,power_law_1.01,0.3710054397583008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,1,32,power_law_1.01,0.42323841094970704
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,1,32,power_law_1.01,0.1945408058166504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,1,32,power_law_1.01,0.5888691329956055
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,1,32,power_law_1.01,0.7136793518066407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,1,32,power_law_1.01,1.0108582305908205
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,1,32,power_law_1.01,1.339411163330078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,1,32,power_law_1.01,2.102170867919922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,1,32,power_law_1.01,0.07857664108276367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,1,32,power_law_1.01,0.07848576068878174
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,1,32,power_law_1.01,0.0760640001296997
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,1,32,power_law_1.01,3.1879064941406248
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,1,32,power_law_1.01,0.07524096012115479
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,1,32,power_law_1.01,0.07046271800994873
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,1,32,power_law_1.01,0.08102527618408203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,1,32,power_law_1.01,0.08221311569213867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,1,32,power_law_1.01,0.0871615982055664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,1,32,power_law_1.01,0.08840191841125489
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,1,32,power_law_1.01,0.09889792442321778
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,1,32,power_law_1.01,4.293140563964844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,1,32,power_law_1.01,0.10434687614440918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,1,32,power_law_1.01,0.10782208442687988
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,1,32,power_law_1.01,0.11196800231933594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,1,32,power_law_1.01,0.11321087837219239
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,1,32,power_law_1.01,0.08946175575256347
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,1,32,power_law_1.01,4.722100524902343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,32,1,power_law_1.2,1.2526060485839845
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,1,32,power_law_1.01,0.12935551643371582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,1,32,power_law_1.01,0.1700953674316406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,1,32,power_law_1.01,0.18145280838012695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,1,32,power_law_1.01,0.23344896316528319
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,1,32,power_law_1.01,0.3051852798461914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,1,32,power_law_1.01,0.37621505737304684
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,1,32,power_law_1.01,0.508482551574707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,1,32,power_law_1.01,0.11699839591979981
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,1,32,power_law_1.01,0.657122573852539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,1,32,power_law_1.2,0.12905088424682618
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,1,32,power_law_1.01,0.9150656127929688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,1,32,power_law_1.2,0.12828800201416016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,1,32,power_law_1.2,0.1277030372619629
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,1,32,power_law_1.2,0.15861632347106933
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,1,32,power_law_1.2,0.15071231842041016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,1,32,power_law_1.2,0.17093503952026368
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,1,32,power_law_1.01,1.6996095275878909
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,1,32,power_law_1.2,0.17897600173950196
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,1,32,power_law_1.01,1.894475555419922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,1,32,power_law_1.2,0.1898624038696289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,1,32,power_law_1.2,0.193255672454834
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,1,32,power_law_1.2,0.19187456130981445
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,1,32,power_law_1.2,0.2029209518432617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,1,32,power_law_1.2,0.214147834777832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,1,32,power_law_1.2,0.23413503646850584
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,1,32,power_law_1.2,0.24052608489990232
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,1,32,power_law_1.2,0.25468671798706055
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,1,32,power_law_1.2,0.2995648002624512
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,1,32,power_law_1.2,0.40798591613769536
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,1,32,power_law_1.2,0.5195942306518555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,1,32,power_law_1.2,0.2042521667480469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,1,32,power_law_1.2,0.714566421508789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,1,32,power_law_1.2,0.9068096160888672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,1,32,power_law_1.2,1.2069631958007814
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,32,1,power_law_1.2,1.1609273529052735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,1,32,power_law_1.2,0.07855616092681886
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,1,32,power_law_1.2,0.07836287975311279
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,1,32,power_law_1.2,2.0800294494628906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,1,32,power_law_1.2,0.07783679962158203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,1,32,power_law_1.2,0.07367167949676515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,1,32,power_law_1.2,0.07280000209808349
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,1,32,power_law_1.2,0.08227840423583985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,1,32,power_law_1.2,0.08096896171569824
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,1,32,power_law_1.2,2.9209368896484373
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,1,32,power_law_1.2,0.08971391677856445
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,1,32,power_law_1.2,0.09024895668029785
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,1,32,power_law_1.2,0.10218239784240724
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,1,32,power_law_1.2,0.09456640243530273
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,1,32,power_law_1.2,0.10473343849182129
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,1,32,power_law_1.2,0.10734975814819336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,1,32,power_law_1.2,0.11399423599243164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,1,32,power_law_1.2,0.11186047554016114
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,1,32,power_law_1.2,0.11319295883178711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,1,32,power_law_1.2,0.13197952270507812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,1,32,power_law_1.2,0.17875711441040038
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,1,32,power_law_1.2,0.20100095748901364
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,1,32,power_law_1.2,0.2736742401123047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,1,32,power_law_1.2,0.34682239532470704
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,1,32,power_law_1.2,0.5229427337646484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,1,32,power_law_1.2,4.2725720214843745
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,1,32,power_law_1.2,0.705354232788086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,1,32,power_law_1.2,1.0256883239746093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,2,16,balanced,0.05295104026794434
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,2,16,balanced,0.054611201286315914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,2,16,balanced,0.06108416080474853
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,1,32,power_law_1.2,1.3105433654785155
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,2,16,balanced,0.07553279876708985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,2,16,balanced,0.11051903724670412
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,2,16,balanced,0.1823539161682129
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,2,16,balanced,0.18777471542358398
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,2,16,balanced,0.18747648239135742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,2,16,balanced,0.18962303161621094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,2,16,balanced,0.1936934471130371
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,2,16,balanced,0.19529983520507813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,1,32,power_law_1.2,7.144777221679687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,2,16,balanced,0.19982336044311524
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,2,16,balanced,0.20088319778442382
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,2,16,balanced,0.20360063552856444
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,2,16,balanced,0.19912704467773437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,2,16,balanced,0.2069248008728027
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,1,32,power_law_1.2,2.434967041015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,2,16,balanced,0.21826431274414065
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,2,16,balanced,0.234967041015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,2,16,balanced,0.2502579116821289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,2,16,balanced,0.33572608947753907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,1,32,power_law_1.2,7.639567260742187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,2,16,balanced,0.3735347366333008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,2,16,balanced,0.5326809692382812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,2,16,balanced,0.6363417434692382
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,2,16,balanced,0.030831360816955568
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,2,16,balanced,0.8201023864746094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,2,16,balanced,0.030894079208374024
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,2,16,balanced,0.038037760257720946
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,2,16,balanced,0.04894207954406739
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,2,16,balanced,1.0848678588867187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,2,16,balanced,0.07173247814178467
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,2,16,balanced,0.09151871681213379
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,2,16,balanced,0.09155839920043946
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,2,16,balanced,0.09327360153198241
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,2,16,balanced,0.09384063720703126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,2,16,balanced,1.5636186218261718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,2,16,balanced,0.09600000381469727
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,2,16,balanced,0.09919360160827637
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,2,16,balanced,0.10131839752197265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,2,16,balanced,0.10667519569396973
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,2,16,balanced,0.09959296226501466
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,2,16,balanced,0.1021401596069336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,2,16,balanced,0.10665599822998047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,2,16,balanced,0.11227264404296874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,2,16,balanced,2.0302374267578123
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,2,16,balanced,0.12624640464782716
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,2,16,balanced,0.13899264335632325
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,2,16,balanced,0.1644416046142578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,2,16,balanced,0.18780160903930665
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,2,16,balanced,0.2673651123046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,2,16,balanced,0.31258880615234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,2,16,balanced,0.4463654327392578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,2,16,balanced,0.5806143951416016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,2,16,power_law_1.01,0.1079526424407959
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,2,16,balanced,0.7733529663085937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,2,16,power_law_1.01,0.13896191596984864
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,2,16,power_law_1.01,0.13830016136169435
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,2,16,power_law_1.01,0.13851263999938965
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,2,16,balanced,1.0398745727539063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,2,16,power_law_1.01,0.1494758415222168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,2,16,power_law_1.01,0.168799991607666
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,2,16,power_law_1.01,0.17524736404418945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,2,16,power_law_1.01,0.18198400497436523
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,2,16,power_law_1.01,0.1910361671447754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,2,16,power_law_1.01,0.19408384323120118
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,2,16,power_law_1.01,0.19703807830810546
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,2,16,power_law_1.01,0.20155775070190432
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,2,16,power_law_1.01,0.19980928421020508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,2,16,power_law_1.01,0.21469951629638673
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,2,16,power_law_1.01,0.2233305549621582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,2,16,power_law_1.01,0.2502668762207031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,2,16,power_law_1.01,0.2592959976196289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,2,16,power_law_1.01,0.3095884895324707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,2,16,power_law_1.01,0.36506622314453124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,2,16,power_law_1.01,0.47741439819335935
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,2,16,power_law_1.01,0.5879308700561523
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,2,16,power_law_1.01,0.8812274932861328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,2,16,power_law_1.01,1.149306869506836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,2,16,power_law_1.01,0.06914303779602052
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,2,16,power_law_1.01,1.5998361206054688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,2,16,power_law_1.01,0.07969024181365966
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,32,1,power_law_1.2,1.6612760925292966
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,2,16,power_law_1.01,0.08072959899902343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,2,16,power_law_1.01,0.07960832118988037
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,2,16,power_law_1.01,0.08418944358825683
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,2,16,power_law_1.01,0.07294847965240478
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,2,16,power_law_1.01,2.1875212097167966
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,2,16,power_law_1.01,0.09072896003723144
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,2,16,power_law_1.01,0.09148544311523438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,2,16,power_law_1.01,0.09626496315002442
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,2,16,power_law_1.01,0.09984895706176758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,2,16,power_law_1.01,0.10598015785217285
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,2,16,power_law_1.01,0.10257280349731446
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,2,16,power_law_1.01,0.10757504463195802
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,2,16,power_law_1.01,3.0682214355468753
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,2,16,power_law_1.01,0.11175935745239259
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,2,16,power_law_1.01,0.0872486400604248
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,2,16,power_law_1.01,0.12120063781738281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,1,32,power_law_1.2,2.7453274536132812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,2,16,power_law_1.01,0.14414591789245607
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,2,16,power_law_1.01,0.18690176010131837
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,2,16,power_law_1.01,0.2409984016418457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,2,16,power_law_1.01,0.2791667175292969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,2,16,power_law_1.01,0.10662528038024903
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,2,16,power_law_1.01,0.3603788757324219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,2,16,power_law_1.01,0.48159744262695314
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,2,16,power_law_1.2,0.10772224426269532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,2,16,power_law_1.01,0.6659648132324218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,2,16,power_law_1.2,0.11199872016906738
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,2,16,power_law_1.01,0.8295526123046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,2,16,power_law_1.2,0.1385689640045166
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,2,16,power_law_1.01,4.474022521972656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,2,16,power_law_1.2,0.12328703880310059
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,2,16,power_law_1.2,0.1517414379119873
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,2,16,power_law_1.2,0.1740991973876953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,2,16,power_law_1.2,0.18144767761230468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,2,16,power_law_1.2,0.18893440246582033
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,2,16,power_law_1.2,0.18589183807373047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,2,16,power_law_1.2,0.1930227279663086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,2,16,power_law_1.01,1.2845721435546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,2,16,power_law_1.2,0.2041843223571777
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,2,16,power_law_1.2,0.20281600952148438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,2,16,power_law_1.2,0.20828800201416015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,2,16,power_law_1.2,0.20919168472290037
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,2,16,power_law_1.2,0.22357503890991212
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,2,16,power_law_1.2,0.23904640197753907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,2,16,power_law_1.2,0.2658880043029785
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,2,16,power_law_1.2,0.3278911972045898
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,2,16,power_law_1.2,0.40794113159179685
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,2,16,power_law_1.2,0.5784204864501954
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,2,16,power_law_1.2,0.6947929382324218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,2,16,power_law_1.2,1.036917724609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,2,16,power_law_1.01,1.6975167846679688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,2,16,power_law_1.2,0.06876927852630615
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,2,16,power_law_1.2,0.0745408010482788
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,2,16,power_law_1.2,0.07608575820922851
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,2,16,power_law_1.2,1.4380685424804687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,2,16,power_law_1.2,0.07412864208221435
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,2,16,power_law_1.2,0.07586304187774659
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,2,16,power_law_1.2,0.08233856201171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,2,16,power_law_1.2,0.08759424209594727
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,2,16,power_law_1.2,0.08949503898620606
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,2,16,power_law_1.2,0.09651455879211426
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,2,16,power_law_1.2,0.10519424438476563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,2,16,power_law_1.2,3.0545230102539063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,32,1,power_law_1.2,1.5167948913574218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,2,16,power_law_1.2,2.1391334533691406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,2,16,power_law_1.2,0.10587648391723632
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,2,16,power_law_1.2,0.10640512466430666
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,2,16,power_law_1.2,0.10614912033081055
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,2,16,power_law_1.2,0.11315327644348146
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,2,16,power_law_1.2,0.12738944053649903
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,2,16,power_law_1.2,0.16007295608520508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,2,16,power_law_1.2,0.20564096450805663
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,2,16,power_law_1.2,0.23836032867431642
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,2,16,power_law_1.2,0.2985958480834961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,2,16,power_law_1.2,0.10826623916625977
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,2,16,power_law_1.2,0.43654144287109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,2,16,power_law_1.2,0.09558143615722656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,2,16,power_law_1.2,4.169427185058593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,4,8,balanced,0.02961663961410522
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,4,8,balanced,0.05267072200775147
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,4,8,balanced,0.059919362068176274
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,2,16,power_law_1.2,0.5834969711303711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,2,16,power_law_1.2,0.8496691131591797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,4,8,balanced,0.07571839809417724
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,4,8,balanced,0.11132927894592286
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,2,16,power_law_1.2,1.101800994873047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,4,8,balanced,0.19230592727661133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,4,8,balanced,0.18152576446533203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,4,8,balanced,0.18712959289550782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,4,8,balanced,0.1970047950744629
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,2,16,power_law_1.2,5.804624633789063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,4,8,balanced,0.19151103973388672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,4,8,balanced,0.19312383651733397
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,4,8,balanced,0.1970419120788574
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,4,8,balanced,0.2033036804199219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,4,8,balanced,0.2190771293640137
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,4,8,balanced,0.22755712509155274
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,4,8,balanced,0.250195198059082
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,4,8,balanced,0.280765438079834
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,4,8,balanced,0.3757823944091797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,4,8,balanced,0.4345868682861328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,4,8,balanced,0.19376768112182616
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,4,8,balanced,0.6302412796020508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,4,8,balanced,0.7593510437011719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,4,8,balanced,0.2107711982727051
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,4,8,balanced,0.029509119987487793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,4,8,balanced,0.031235840320587162
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,4,8,balanced,1.0140287780761719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,4,8,balanced,0.03798784017562866
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,2,16,power_law_1.2,1.6503884887695315
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,4,8,balanced,0.05236991882324219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,4,8,balanced,0.07394303798675536
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,4,8,balanced,0.09682687759399414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,4,8,balanced,1.333763885498047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,4,8,balanced,0.10012928009033203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,4,8,balanced,0.10251008033752443
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,4,8,balanced,0.10567680358886719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,4,8,balanced,0.10682496070861816
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,4,8,balanced,0.10123519897460938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,4,8,balanced,0.10450431823730469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,4,8,balanced,0.11188735961914062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,4,8,balanced,0.1164031982421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,4,8,balanced,0.12371583938598633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,4,8,balanced,0.1347059154510498
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,4,8,balanced,0.15739007949829104
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,4,8,balanced,0.17812992095947267
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,4,8,balanced,1.9329087829589846
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,4,8,balanced,0.21836544036865235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,4,8,balanced,0.25906431198120117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,4,8,balanced,0.3783027267456055
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,4,8,balanced,0.46240127563476563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,4,8,balanced,2.539654388427734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,4,8,balanced,0.6669631958007812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,4,8,power_law_1.01,0.07352960109710693
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,4,8,power_law_1.01,0.09184639930725098
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,4,8,balanced,0.8732621002197266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,4,8,power_law_1.01,0.08802176475524902
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,4,8,power_law_1.01,0.1159385585784912
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,4,8,power_law_1.01,0.1158233642578125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,4,8,power_law_1.01,0.17732736587524414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,4,8,power_law_1.01,0.18049535751342774
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,4,8,balanced,1.2115545654296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,4,8,power_law_1.01,0.19077375411987305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,4,8,power_law_1.01,0.19617919921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,4,8,power_law_1.01,0.19590784072875977
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,4,8,power_law_1.01,0.19228416442871094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,4,8,balanced,0.1078003215789795
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,4,8,balanced,1.6231295776367187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,4,8,power_law_1.01,0.2072742462158203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,4,8,power_law_1.01,0.21729536056518556
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,4,8,power_law_1.01,0.22742528915405275
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,4,8,power_law_1.01,0.27246976852416993
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,4,8,power_law_1.01,0.1705023956298828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,4,8,power_law_1.01,0.33473022460937496
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,4,8,power_law_1.01,0.4254272079467773
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,4,8,power_law_1.01,0.1943155288696289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,4,8,power_law_1.01,0.5471513748168946
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,4,8,power_law_1.01,0.19651968002319337
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,4,8,power_law_1.01,0.7050982666015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,4,8,power_law_1.01,0.8689356994628905
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,4,8,power_law_1.01,0.04982399940490723
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,4,8,power_law_1.01,0.060323839187622075
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,4,8,power_law_1.01,1.2430528259277343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,4,8,power_law_1.01,0.05785215854644775
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,4,8,power_law_1.01,0.07092351913452148
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,4,8,power_law_1.01,0.07440896034240722
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,4,8,power_law_1.01,0.08918016433715821
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,4,8,power_law_1.01,0.09186047554016114
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,4,8,power_law_1.01,1.683118133544922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,4,8,power_law_1.01,0.0951142406463623
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,4,8,power_law_1.01,0.10088191986083986
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,4,8,power_law_1.01,0.10930432319641112
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,4,8,power_law_1.01,0.10103808403015138
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,4,8,power_law_1.01,0.10642944335937501
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,4,8,power_law_1.01,2.390442199707031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,4,8,power_law_1.01,0.10969216346740722
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,4,8,power_law_1.01,0.1131276798248291
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,4,8,power_law_1.01,0.13085056304931642
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,4,8,power_law_1.01,0.14484864234924316
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,4,8,power_law_1.01,0.0969536018371582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,4,8,power_law_1.01,0.1715750312805176
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,4,8,power_law_1.01,0.10738431930541992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,4,8,power_law_1.01,0.27956096649169926
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,4,8,power_law_1.01,0.3790630340576172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,4,8,power_law_1.01,3.1345062255859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,4,8,power_law_1.01,0.47918846130371096
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,4,8,power_law_1.01,0.6057638549804688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,4,8,power_law_1.2,0.08123519897460937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,4,8,power_law_1.2,0.07633920192718506
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,4,8,power_law_1.2,0.09113727569580078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,4,8,power_law_1.2,0.10294015884399414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,4,8,power_law_1.01,0.8281664276123047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,4,8,power_law_1.2,0.10411264419555663
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,4,8,power_law_1.2,0.17115903854370118
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,4,8,power_law_1.01,0.23515520095825196
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,4,8,power_law_1.2,0.17408128738403322
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,4,8,power_law_1.2,0.1781657600402832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,4,8,power_law_1.2,0.19392383575439454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,4,8,power_law_1.01,1.279727325439453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,4,8,power_law_1.2,0.19235071182250976
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,4,8,power_law_1.2,0.1933145523071289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,4,8,power_law_1.2,0.19347200393676758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,4,8,power_law_1.2,0.19694335937500002
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,4,8,power_law_1.2,0.219998722076416
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,4,8,power_law_1.2,0.22866943359374997
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,4,8,power_law_1.2,0.20509311676025394
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,4,8,power_law_1.2,0.3000294494628906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,4,8,power_law_1.01,1.8004135131835937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,4,8,power_law_1.2,0.35080062866210937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,4,8,power_law_1.2,0.47187328338623047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,4,8,power_law_1.2,0.1895884895324707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,4,8,power_law_1.2,0.5702143859863281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,4,8,power_law_1.2,0.7330355072021485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,4,8,power_law_1.2,0.04946047782897949
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,4,8,power_law_1.2,0.05090303897857666
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,4,8,power_law_1.2,0.05904128074645996
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,4,8,power_law_1.2,0.06541567802429199
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,4,8,power_law_1.2,0.07189119815826416
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,4,8,power_law_1.2,1.368775634765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,4,8,power_law_1.2,0.08921600341796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,4,8,power_law_1.2,0.09180543899536134
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,4,8,power_law_1.2,0.09835264205932617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,4,8,power_law_1.2,0.09698304176330566
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,2,16,power_law_1.2,2.2005567932128907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,4,8,power_law_1.2,0.10073472023010255
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,4,8,power_law_1.2,1.96044677734375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,4,8,power_law_1.2,0.09967103958129883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,4,8,power_law_1.2,1.013854751586914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,4,8,power_law_1.2,0.10355199813842772
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,4,8,power_law_1.2,0.10524800300598144
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,4,8,power_law_1.2,0.1111308765411377
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,4,8,power_law_1.2,0.11570688247680665
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,4,8,power_law_1.2,0.12570879936218263
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,4,8,power_law_1.2,0.15884672164916994
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,4,8,power_law_1.2,0.19379072189331054
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,4,8,power_law_1.2,0.2519398307800293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,4,8,power_law_1.2,0.2779827117919922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,4,8,power_law_1.2,0.39967487335205076
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,4,8,power_law_1.2,0.5087385559082032
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,4,8,power_law_1.2,2.973576965332031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,4,8,power_law_1.2,0.10594047546386717
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,8,4,balanced,0.03010175943374634
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,8,4,balanced,0.052275199890136716
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,8,4,balanced,0.06083072185516357
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,4,8,power_law_1.2,0.720847396850586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,8,4,balanced,0.07654016017913819
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,8,4,balanced,0.11313152313232422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,8,4,balanced,0.13669247627258302
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,8,4,balanced,0.18457855224609374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,8,4,balanced,0.18594688415527344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,8,4,balanced,0.19011199951171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,8,4,balanced,0.19068416595458984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,8,4,balanced,0.1856332778930664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,8,4,balanced,0.18722431182861327
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,8,4,balanced,0.19217792510986328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,8,4,balanced,0.19622655868530275
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,8,4,balanced,0.20137983322143554
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,8,4,balanced,0.20787967681884764
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,8,4,balanced,0.21826175689697264
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,8,4,balanced,0.23918336868286133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,8,4,balanced,0.25712511062622073
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,8,4,balanced,0.36929534912109374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,8,4,balanced,0.4259135818481445
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,8,4,balanced,0.6088614273071289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,4,8,power_law_1.2,1.0031731414794922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,8,4,balanced,0.7584780883789063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,4,8,power_law_1.2,3.827785034179688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,8,4,balanced,1.0120626831054689
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,4,8,power_law_1.2,1.4560269165039064
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,8,4,balanced,0.03035264015197754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,8,4,balanced,0.032440319061279296
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,8,4,balanced,0.04356351852416992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,8,4,balanced,0.05592319965362549
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,8,4,balanced,0.08210687637329102
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,8,4,balanced,0.09797887802124024
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,8,4,balanced,1.3122483825683593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,8,4,balanced,0.11819775581359863
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,8,4,balanced,0.12046848297119142
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,8,4,balanced,0.12219264030456542
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,8,4,balanced,0.11536383628845215
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,8,4,balanced,0.11851648330688476
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,8,4,balanced,0.11981951713562011
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,8,4,balanced,0.12460160255432129
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,8,4,balanced,0.12980735778808594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,8,4,balanced,0.134398717880249
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,8,4,balanced,0.14451711654663085
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,8,4,balanced,0.16522111892700195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,8,4,balanced,0.1826150321960449
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,8,4,balanced,0.22263679504394532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,8,4,balanced,0.11637248039245604
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,8,4,balanced,0.26539520263671873
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,8,4,balanced,0.3930022430419922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,8,4,balanced,2.491327972412109
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,8,4,balanced,0.4712115097045898
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,8,4,power_law_1.01,0.06063360214233399
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,8,4,balanced,0.6795597076416016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,8,4,power_law_1.01,0.07321856021881104
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,8,4,balanced,1.903459777832031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,8,4,balanced,0.8929663848876952
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,8,4,power_law_1.01,0.08750335693359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,8,4,power_law_1.01,0.09573887825012208
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,8,4,power_law_1.01,0.12534144401550293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,8,4,power_law_1.01,0.17062911987304688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,8,4,power_law_1.01,0.17896831512451172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,8,4,balanced,1.2631398773193359
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,8,4,power_law_1.01,0.18230911254882814
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,8,4,power_law_1.01,0.18202367782592774
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,8,4,power_law_1.01,0.18582271575927733
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,8,4,balanced,1.683251190185547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,8,4,power_law_1.01,0.19243135452270507
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,8,4,power_law_1.01,0.20408704757690427
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,8,4,power_law_1.01,0.07613440036773682
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,8,4,power_law_1.01,0.21439231872558592
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,8,4,power_law_1.01,0.2472819137573242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,8,4,power_law_1.01,0.17266815185546874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,8,4,power_law_1.01,0.2894668769836426
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,8,4,power_law_1.01,0.3803251266479492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,8,4,power_law_1.01,0.18328191757202147
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,8,4,power_law_1.01,0.476627197265625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,8,4,power_law_1.01,0.6433318328857422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,8,4,power_law_1.01,0.18930944442749023
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,8,4,power_law_1.01,0.8297881317138671
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,8,4,power_law_1.01,0.04868735790252686
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,8,4,power_law_1.01,1.1272000122070314
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,8,4,power_law_1.01,0.05409279823303222
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,8,4,power_law_1.01,0.06372352123260498
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,8,4,power_law_1.01,0.06927103996276855
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,8,4,power_law_1.01,0.10582143783569335
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,8,4,power_law_1.01,1.55538818359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,8,4,power_law_1.01,0.10884480476379395
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,8,4,power_law_1.01,0.11499263763427733
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,8,4,power_law_1.01,0.04173567771911621
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,8,4,power_law_1.01,0.12030591964721679
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,4,8,power_law_1.2,1.826945343017578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,8,4,power_law_1.01,0.11196415901184083
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,8,4,power_law_1.01,0.11626624107360839
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,8,4,power_law_1.01,0.12047743797302246
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,8,4,power_law_1.01,2.175722198486328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,8,4,power_law_1.01,0.12300288200378418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,8,4,power_law_1.01,0.12747391700744629
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,8,4,power_law_1.01,0.14018176078796388
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,8,4,power_law_1.01,0.08516736030578613
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,8,4,power_law_1.01,0.17685375213623047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,8,4,power_law_1.01,0.2122188758850098
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,8,4,power_law_1.01,0.26007808685302736
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,8,4,power_law_1.01,0.11175807952880859
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,8,4,power_law_1.01,0.30620288848876953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,8,4,power_law_1.01,2.7680538940429686
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,8,4,power_law_1.01,0.4073292922973633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,8,4,power_law_1.2,0.05989888191223145
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,8,4,power_law_1.01,0.5232998275756836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,8,4,power_law_1.2,0.06436480045318603
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,8,4,power_law_1.2,0.07229055881500243
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,8,4,power_law_1.01,0.7478899383544922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,8,4,power_law_1.2,0.07793024063110351
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,8,4,power_law_1.2,0.0969651222229004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,8,4,power_law_1.2,0.11646207809448242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,8,4,power_law_1.2,0.1675904083251953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,8,4,power_law_1.01,0.9602547454833985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,8,4,power_law_1.2,0.17528703689575195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,8,4,power_law_1.2,0.18538240432739256
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,8,4,power_law_1.2,0.18797439575195313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,8,4,power_law_1.2,0.1858073616027832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,8,4,power_law_1.2,0.18081792831420898
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,8,4,power_law_1.2,0.19057920455932617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,8,4,power_law_1.01,1.3304165649414064
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,8,4,power_law_1.2,0.19928192138671874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,8,4,power_law_1.2,0.20603136062622068
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,8,4,power_law_1.2,0.21733119964599606
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,8,4,power_law_1.2,0.256254711151123
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,8,4,power_law_1.01,1.7770649719238283
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,8,4,power_law_1.2,0.31175167083740235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,8,4,power_law_1.2,0.412933120727539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,8,4,power_law_1.2,0.17918079376220702
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,8,4,power_law_1.2,0.48313217163085936
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,8,4,power_law_1.2,0.6694566345214843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,8,4,power_law_1.2,0.04146815776824951
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,8,4,power_law_1.2,0.896468505859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,8,4,power_law_1.2,0.045203199386596685
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,8,4,power_law_1.2,0.051201281547546384
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,8,4,power_law_1.2,0.05609216213226318
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,8,4,power_law_1.2,1.2416627502441406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,8,4,power_law_1.2,0.10498944282531739
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,8,4,power_law_1.2,0.10903039932250977
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,8,4,power_law_1.2,0.11237248420715332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,8,4,power_law_1.2,0.11541248321533204
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,8,4,power_law_1.2,0.11605376243591307
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,8,4,power_law_1.2,0.1128883171081543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,8,4,power_law_1.2,0.11553791999816894
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,8,4,power_law_1.2,0.11959936141967772
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,8,4,power_law_1.2,0.12583807945251463
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,8,4,power_law_1.2,0.13364480018615724
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,8,4,power_law_1.2,0.15242624282836914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,8,4,power_law_1.2,0.18498176574707031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,8,4,power_law_1.2,0.21779327392578124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,8,4,power_law_1.2,0.2667110443115234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,8,4,power_law_1.2,0.31734399795532225
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,8,4,power_law_1.2,0.43630081176757807
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,8,4,power_law_1.2,0.5364006423950196
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,8,4,power_law_1.2,1.5345497131347656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,8,4,power_law_1.2,0.7718374633789062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,16,2,balanced,0.03680255889892578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,16,2,balanced,0.05280767917633057
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,16,2,balanced,0.06252287864685059
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,8,4,power_law_1.2,3.211635131835938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,8,4,power_law_1.2,2.3301913452148435
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,8,4,power_law_1.2,0.07043712139129639
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,16,2,balanced,0.12079872131347655
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,8,4,power_law_1.2,0.08475775718688965
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,16,2,balanced,0.14749183654785156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,16,2,balanced,0.15019264221191406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,16,2,balanced,0.15092096328735352
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,16,2,balanced,0.18157951354980467
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,16,2,balanced,0.18602880477905273
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,8,4,power_law_1.2,0.9737945556640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,16,2,balanced,0.18959871292114258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,16,2,balanced,0.19563392639160154
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,16,2,balanced,0.19780351638793944
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,16,2,balanced,0.20501119613647462
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,8,4,power_law_1.2,1.4621888732910155
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,16,2,balanced,0.08214783668518066
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,16,2,balanced,0.21440383911132815
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,16,2,balanced,0.23250816345214842
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,16,2,balanced,0.25038848876953124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,16,2,balanced,0.1460927963256836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,16,2,balanced,0.3724620819091797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,16,2,balanced,0.41691135406494145
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,16,2,balanced,0.6012723159790039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,16,2,balanced,0.18732799530029295
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,16,2,balanced,0.030378239154815672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,16,2,balanced,0.7664921569824219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,16,2,balanced,0.03398783922195435
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,16,2,balanced,0.04444799900054931
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,8,4,power_law_1.2,1.934668731689453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,16,2,balanced,0.06434688091278076
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,16,2,balanced,0.09899776458740235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,16,2,balanced,0.11882880210876465
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,16,2,balanced,0.11900544166564941
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,16,2,balanced,0.12155136108398437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,16,2,balanced,1.3427008056640626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,16,2,balanced,0.17461631774902345
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,16,2,balanced,0.17441791534423828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,16,2,balanced,0.17825279235839844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,16,2,balanced,0.1804377555847168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,16,2,balanced,0.18537984848022462
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,16,2,balanced,0.18722816467285158
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,16,2,balanced,0.1927987289428711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,16,2,balanced,0.20028160095214842
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,16,2,balanced,1.9306906127929686
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,16,2,balanced,0.23403263092041016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,16,2,balanced,0.2719027137756348
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,16,2,balanced,1.0234444427490235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,16,2,balanced,0.30835327148437497
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,16,2,balanced,0.11845760345458985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,16,2,balanced,0.46669055938720705
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,16,2,balanced,2.5161907958984377
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,16,2,balanced,0.5379916763305663
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,16,2,power_law_1.01,0.06002816200256348
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,16,2,balanced,0.7770304107666016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,16,2,balanced,0.21986431121826172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,16,2,balanced,1.0079270172119141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,16,2,power_law_1.01,0.06862463951110839
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,16,2,power_law_1.01,0.08012288093566895
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,16,2,power_law_1.01,0.09349760055541992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,16,2,power_law_1.01,0.12874367713928223
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,16,2,power_law_1.01,0.12210816383361815
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,16,2,power_law_1.01,0.13317503929138183
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,16,2,power_law_1.01,0.1406873607635498
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,16,2,balanced,1.4247514343261718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,16,2,power_law_1.01,0.1732377624511719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,16,2,power_law_1.01,0.1763199996948242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,16,2,power_law_1.01,0.18191999435424805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,16,2,power_law_1.01,0.1829555130004883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,16,2,power_law_1.01,0.19461759567260742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,16,2,power_law_1.01,0.18844160079956054
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,16,2,power_law_1.01,0.20235904693603515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,16,2,power_law_1.01,0.21468416213989258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,16,2,balanced,1.8888461303710937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,16,2,power_law_1.01,0.24149248123168948
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,16,2,power_law_1.01,0.26751232147216797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,16,2,power_law_1.01,0.3689318466186523
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,16,2,power_law_1.01,0.4363225555419922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,16,2,power_law_1.01,0.621077766418457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,16,2,power_law_1.01,0.7487347412109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,16,2,power_law_1.01,0.032934401035308834
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,16,2,power_law_1.01,0.04258815765380859
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,16,2,power_law_1.01,0.050079998970031736
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,16,2,power_law_1.01,1.0641126251220703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,16,2,power_law_1.01,0.061967358589172364
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,16,2,power_law_1.01,0.0765721607208252
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,16,2,power_law_1.01,0.1028940773010254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,16,2,power_law_1.01,0.05284224033355713
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,16,2,power_law_1.01,1.3799334716796874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,16,2,power_law_1.01,0.10674688339233399
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,16,2,power_law_1.01,0.10961536407470703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,16,2,power_law_1.01,0.16282495498657226
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,16,2,power_law_1.01,0.16864000320434572
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,16,2,power_law_1.01,0.1744793510437012
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,16,2,power_law_1.01,0.17523584365844727
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,16,2,power_law_1.01,0.1796659278869629
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,16,2,power_law_1.01,0.18393600463867188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,16,2,power_law_1.01,0.18790016174316407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,16,2,power_law_1.01,2.0274855041503903
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,16,2,power_law_1.01,0.20186111450195315
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,16,2,power_law_1.01,0.2181644821166992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,16,2,power_law_1.01,0.24165632247924806
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,16,2,power_law_1.01,0.29217664718627934
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,16,2,power_law_1.01,0.3491545486450195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,16,2,power_law_1.01,0.09914112091064453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,16,2,power_law_1.01,0.4757452774047851
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,16,2,power_law_1.01,2.649811096191406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,16,2,power_law_1.2,0.0531763219833374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,16,2,power_law_1.01,0.5779059219360352
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,16,2,power_law_1.2,0.05717887878417969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,16,2,power_law_1.2,0.0684339189529419
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,16,2,power_law_1.2,0.08887807846069336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,16,2,power_law_1.01,0.8331302642822266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,16,2,power_law_1.2,0.12980735778808594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,16,2,power_law_1.01,1.0362163543701173
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,16,2,power_law_1.2,0.1385536003112793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,16,2,power_law_1.2,0.17728639602661134
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,16,2,power_law_1.2,0.17833343505859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,16,2,power_law_1.2,0.18256895065307616
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,16,2,power_law_1.01,1.4883993530273438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,16,2,power_law_1.2,0.18321407318115235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,16,2,power_law_1.2,0.1898918342590332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,16,2,power_law_1.2,0.1966681671142578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,16,2,power_law_1.2,0.07568511962890626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,16,2,power_law_1.2,0.21766784667968747
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,16,2,power_law_1.2,0.2045734405517578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,16,2,power_law_1.2,0.24734975814819338
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,16,2,power_law_1.2,0.2753523254394531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,16,2,power_law_1.2,0.13668864250183105
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,16,2,power_law_1.2,0.3796121597290039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,16,2,power_law_1.2,0.44571647644042967
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,16,2,power_law_1.2,0.6344588851928711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,16,2,power_law_1.2,0.11995136260986328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,16,2,power_law_1.2,0.7907148742675781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,16,2,power_law_1.2,0.04064511775970459
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,16,2,power_law_1.2,0.05255296230316162
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,16,2,power_law_1.2,0.05933824062347413
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,16,2,power_law_1.2,0.07165567874908448
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,16,2,power_law_1.2,1.433471984863281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,16,2,power_law_1.2,0.0990015983581543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,16,2,power_law_1.2,0.10315648078918457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,16,2,power_law_1.2,0.10886783599853515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,16,2,power_law_1.2,0.10857728004455566
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,16,2,power_law_1.2,0.033569281101226804
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,16,2,power_law_1.2,0.16527999877929686
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,16,2,power_law_1.2,0.17088640213012696
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,16,2,power_law_1.2,0.1730611228942871
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,16,2,power_law_1.01,1.9869389343261719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,16,2,power_law_1.2,0.17554304122924805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,16,2,power_law_1.2,0.17300991058349607
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,16,2,power_law_1.2,0.18521728515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,16,2,power_law_1.2,0.1913408088684082
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,16,2,power_law_1.2,1.0833715057373046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,16,2,power_law_1.2,0.2259187126159668
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,16,2,power_law_1.2,0.24780927658081056
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,16,2,power_law_1.2,2.115120697021484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,16,2,power_law_1.2,0.297446403503418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,16,2,power_law_1.2,0.3662604904174805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,16,2,power_law_1.2,0.48134014129638675
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,16,2,power_law_1.2,0.5838528060913086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,32,1,balanced,0.04442815780639649
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,16,2,power_law_1.2,0.20633216857910158
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,32,1,balanced,0.0652569580078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,32,1,balanced,0.07787199974060058
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,32,1,balanced,0.10286591529846192
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,32,1,balanced,0.15286335945129395
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,16,2,power_law_1.2,2.7385919189453127
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,32,1,balanced,0.18681472778320313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,32,1,balanced,0.18398080825805666
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,32,1,balanced,0.18662975311279298
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,32,1,balanced,0.18731456756591797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,32,1,balanced,0.19066560745239258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,32,1,balanced,0.19779455184936523
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,32,1,balanced,0.20095104217529297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,16,2,power_law_1.2,0.8264575958251953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,16,2,power_law_1.2,1.0782310485839843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,32,1,balanced,0.22022464752197263
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,32,1,balanced,0.22403072357177733
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,16,2,power_law_1.2,1.5207244873046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,16,2,power_law_1.2,2.067530212402344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,32,1,balanced,0.2396806335449219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,32,1,balanced,0.2616998481750488
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,32,1,balanced,0.28414976119995117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,32,1,balanced,0.45128513336181636
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,32,1,balanced,0.6541305541992187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,32,1,balanced,0.22545856475830078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,32,1,balanced,0.8817721557617186
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,32,1,balanced,0.04153600215911865
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,32,1,balanced,0.050424318313598636
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,32,1,balanced,0.06477824211120606
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,32,1,balanced,0.0915167999267578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,32,1,balanced,0.14861056327819824
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,32,1,balanced,0.23152320861816406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,32,1,balanced,0.1910598373413086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,32,1,balanced,1.5733407592773436
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,32,1,balanced,0.1906559944152832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,32,1,balanced,0.19118719100952147
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,32,1,balanced,0.1935763168334961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,32,1,balanced,0.19571775436401367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,32,1,balanced,0.1960691261291504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,32,1,balanced,1.1602444458007812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,32,1,balanced,0.19854335784912108
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,32,1,balanced,0.24636352539062498
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,32,1,balanced,0.24957183837890623
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,32,1,balanced,0.2500441551208496
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,32,1,balanced,0.25376895904541014
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,32,1,balanced,0.2609849548339844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,32,1,balanced,2.2717797851562502
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,32,1,balanced,0.27982463836669924
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,32,1,balanced,0.2997599983215332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,32,1,balanced,0.33471553802490234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,32,1,balanced,0.37147327423095705
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,32,1,balanced,0.47757953643798834
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,32,1,balanced,0.5885996627807617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,32,1,balanced,0.6565773010253906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,32,1,power_law_1.01,0.06604351997375488
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,32,1,balanced,0.9508927917480469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,32,1,balanced,2.969281921386719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,32,1,power_law_1.01,0.07795455932617187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,32,1,power_law_1.01,0.09419903755187989
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,32,1,balanced,1.2414694213867188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,32,1,power_law_1.01,0.11701696395874024
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,32,1,power_law_1.01,0.16252096176147462
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,32,1,power_law_1.01,0.15686400413513185
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,32,1,power_law_1.01,0.16808767318725587
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,32,1,power_law_1.01,0.17589567184448243
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,32,1,power_law_1.01,0.18061567306518556
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,32,1,power_law_1.01,0.19312320709228517
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,32,1,power_law_1.01,0.2050726318359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,32,1,balanced,1.704462127685547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,32,1,power_law_1.01,0.21088703155517577
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,32,1,power_law_1.01,0.043529601097106935
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,32,1,power_law_1.01,0.21963647842407225
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,32,1,power_law_1.01,0.223374080657959
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,32,1,power_law_1.01,0.2278233528137207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,32,1,power_law_1.01,0.2437286376953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,32,1,power_law_1.01,0.27412864685058597
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,32,1,power_law_1.01,0.3073388862609863
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,32,1,balanced,2.2831546020507814
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,32,1,power_law_1.01,0.5133491134643554
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,32,1,power_law_1.01,0.6793011474609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,32,1,power_law_1.01,0.04118080139160156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,32,1,power_law_1.01,0.8462553405761719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,32,1,power_law_1.01,0.04944511890411377
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,32,1,power_law_1.01,0.06407936096191406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,32,1,power_law_1.01,1.200464630126953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,32,1,power_law_1.01,0.08309632301330566
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,32,1,power_law_1.01,0.39148094177246096
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,32,1,power_law_1.01,0.10604479789733887
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,32,1,power_law_1.01,0.14485695838928223
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,32,1,power_law_1.01,0.15697983741760252
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,32,1,power_law_1.01,0.16769472122192383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,32,1,power_law_1.01,0.17544448852539063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,32,1,power_law_1.01,1.5704083251953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,32,1,power_law_1.01,0.1788876724243164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,32,1,power_law_1.01,0.18381183624267577
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,32,1,power_law_1.01,0.18999168395996094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,32,1,power_law_1.01,0.2353932762145996
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,32,1,power_law_1.01,0.24697664260864255
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,32,1,power_law_1.01,0.24833471298217774
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,32,1,power_law_1.01,0.25042688369750976
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,32,1,power_law_1.01,0.26352767944335936
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,32,1,power_law_1.01,0.29410432815551757
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,32,1,power_law_1.01,2.296356506347656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,32,1,power_law_1.01,0.32095870971679685
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,32,1,power_law_1.01,0.37454208374023434
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,32,1,power_law_1.01,0.43980224609374996
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,32,1,power_law_1.01,0.5858297729492188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,32,1,power_law_1.01,0.698108139038086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,32,1,power_law_1.2,0.04085504055023194
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,32,1,power_law_1.01,3.019379272460937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,32,1,power_law_1.2,0.06574719905853271
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,32,1,power_law_1.01,0.9768498992919922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,32,1,power_law_1.2,0.07778560161590577
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,32,1,power_law_1.2,0.09023679733276367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,32,1,power_law_1.2,0.1124403190612793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,32,1,power_law_1.2,0.16152320861816408
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,32,1,power_law_1.2,0.1557363224029541
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,32,1,power_law_1.2,0.17001535415649413
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,32,1,power_law_1.2,0.17737791061401367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,32,1,power_law_1.2,0.1856595230102539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,32,1,power_law_1.2,0.1975008010864258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,32,1,power_law_1.2,0.20797376632690429
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,32,1,power_law_1.2,0.21586559295654295
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,32,1,power_law_1.01,1.7698655700683594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,32,1,power_law_1.2,0.2215251159667969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,32,1,power_law_1.2,0.22868480682373046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,32,1,power_law_1.2,0.2364108848571777
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,32,1,power_law_1.01,1.2392262268066407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,32,1,power_law_1.2,0.2774899291992187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,32,1,power_law_1.2,0.31202816009521483
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,32,1,power_law_1.2,0.4146758270263672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,32,1,power_law_1.01,2.305377960205078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,32,1,power_law_1.2,0.5208607864379883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,32,1,power_law_1.2,0.6915167999267579
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,32,1,power_law_1.2,0.04182655811309814
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,32,1,power_law_1.2,0.24856639862060548
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,32,1,power_law_1.2,0.8559200286865234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,32,1,power_law_1.2,0.051146240234374994
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,32,1,power_law_1.2,0.06389056205749512
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,32,1,power_law_1.2,0.07802048206329346
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,32,1,power_law_1.2,0.10300928115844728
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,32,1,power_law_1.2,0.153502721786499
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,32,1,power_law_1.2,1.2116448211669923
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,32,1,power_law_1.2,0.16884544372558594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,32,1,power_law_1.2,0.17700159072875976
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,32,1,power_law_1.2,0.1834079933166504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,32,1,power_law_1.2,0.1877542304992676
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,32,1,power_law_1.2,0.1918534469604492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,32,1,power_law_1.2,0.241713924407959
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,32,1,power_law_1.2,0.24804800033569335
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,32,1,power_law_1.2,0.2540947151184082
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,32,1,power_law_1.2,0.2602828788757324
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,32,1,power_law_1.2,0.1438425636291504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,32,1,power_law_1.2,0.2722035217285156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,32,1,power_law_1.2,1.5844985961914062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,32,1,power_law_1.2,0.29933887481689453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,32,1,power_law_1.2,0.3264870452880859
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,32,1,power_law_1.2,0.38213951110839844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,32,1,power_law_1.2,0.4446156692504883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,32,1,power_law_1.2,0.5937152099609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,32,1,power_law_1.2,0.7120313262939453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,1,32,balanced,0.06839424133300781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,1,32,balanced,0.06868607997894287
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,1,32,balanced,0.06822912216186523
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,1,32,balanced,0.07901823997497559
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,1,32,balanced,0.10863231658935546
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,32,1,power_law_1.2,0.9838848114013672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,1,32,balanced,0.20065919876098634
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,1,32,balanced,0.27329151153564457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,1,32,balanced,0.3880575942993164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,1,32,balanced,0.39133697509765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,1,32,balanced,0.3920537567138672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,1,32,balanced,0.3909414291381836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,1,32,balanced,0.4104460906982422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,32,1,power_law_1.2,1.2459142303466797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,1,32,balanced,0.4061849594116211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,1,32,balanced,0.420711669921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,1,32,balanced,0.41989761352539057
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,1,32,balanced,0.4220620727539062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,1,32,balanced,0.42433792114257807
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,1,32,balanced,0.43174911499023433
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,1,32,balanced,0.4575270462036133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,1,32,balanced,0.48850177764892583
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,1,32,balanced,0.5237734222412109
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,1,32,balanced,0.5820979309082032
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,1,32,balanced,0.6911692810058594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,1,32,balanced,0.756833267211914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,1,32,balanced,0.04062335968017578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,1,32,balanced,1.0231743621826173
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,1,32,balanced,0.04055168151855469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,1,32,balanced,0.04129280090332031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,1,32,balanced,0.04938240051269531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,1,32,balanced,0.06546944141387939
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,32,1,power_law_1.2,2.3153990173339842
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,1,32,balanced,1.3398130798339845
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,1,32,balanced,0.10756352424621582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,1,32,balanced,0.15377535820007324
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,1,32,balanced,0.17091455459594726
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,1,32,balanced,0.17052032470703124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,1,32,balanced,0.16951936721801758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,1,32,balanced,0.17533056259155272
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,1,32,balanced,1.8975424194335937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,1,32,balanced,0.17796096801757813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,1,32,balanced,0.1767705535888672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,1,32,balanced,0.1831923294067383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,1,32,balanced,0.1868390464782715
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,1,32,balanced,0.18777215957641602
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,1,32,balanced,0.18187776565551758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,1,32,balanced,0.1895257568359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,1,32,balanced,0.19834112167358398
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,1,32,balanced,0.21203071594238282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,1,32,balanced,0.23376384735107422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,1,32,balanced,0.26154624938964843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,1,32,balanced,0.2917056083679199
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,1,32,balanced,0.34863105773925784
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,32,1,power_law_1.2,1.7809286499023436
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,1,32,balanced,0.4516787338256836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,1,32,power_law_1.01,0.20174463272094725
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,1,32,power_law_1.01,0.2664204788208008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,1,32,balanced,0.7751667022705078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,1,32,power_law_1.01,0.2677235221862793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,1,32,power_law_1.01,0.2690700721740723
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,1,32,power_law_1.01,0.2673011207580566
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,1,32,power_law_1.01,0.25520383834838867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,1,32,power_law_1.01,0.3652876663208008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,1,32,power_law_1.01,0.359818229675293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,1,32,power_law_1.01,0.36202369689941405
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,1,32,balanced,0.5732185745239258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,1,32,power_law_1.01,0.37414657592773437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,1,32,power_law_1.01,0.3352243041992188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,1,32,power_law_1.01,0.3190937614440918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,1,32,power_law_1.01,0.3266201782226562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,1,32,power_law_1.01,0.35262977600097656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,1,32,power_law_1.01,0.3650265502929687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,1,32,power_law_1.01,0.39998207092285154
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,1,32,power_law_1.01,0.41620735168457035
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,1,32,power_law_1.01,0.47630847930908204
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,1,32,power_law_1.01,0.5162649536132813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,1,32,power_law_1.01,0.6260236740112305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,1,32,power_law_1.01,0.6904908752441405
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,1,32,power_law_1.01,0.8708761596679688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,1,32,power_law_1.01,0.10627072334289552
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,1,32,power_law_1.01,1.1756851196289062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,1,32,power_law_1.01,0.15146495819091796
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,1,32,power_law_1.01,0.15366911888122559
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,1,32,power_law_1.01,1.5930189514160156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,1,32,power_law_1.01,0.15308799743652343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,1,32,power_law_1.01,0.15606783866882323
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,1,32,power_law_1.01,2.038702087402344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,1,32,power_law_1.01,0.16625919342041015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,1,32,power_law_1.01,0.12225919723510743
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,1,32,power_law_1.01,0.15331456184387207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,1,32,power_law_1.01,0.17014015197753907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,1,32,power_law_1.01,0.17130495071411134
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,1,32,power_law_1.01,0.172043514251709
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,1,32,power_law_1.01,0.1763430404663086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,1,32,power_law_1.01,2.8670950317382813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,1,32,power_law_1.01,0.1814182472229004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,1,32,power_law_1.01,0.18103040695190428
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,1,32,power_law_1.01,0.17927295684814454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,1,32,power_law_1.01,0.1882943916320801
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,1,32,power_law_1.01,0.19640832901000976
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,1,32,power_law_1.01,0.21485824584960939
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,1,32,power_law_1.01,0.2447667121887207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,1,32,power_law_1.01,0.29498111724853515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,32,1,power_law_1.2,3.037789306640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,1,32,power_law_1.01,0.175228157043457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,1,32,power_law_1.01,0.43956737518310546
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,1,32,power_law_1.01,4.114878845214844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,1,32,power_law_1.01,0.6211328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,1,32,power_law_1.2,0.19981311798095702
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,1,32,power_law_1.01,0.7991449737548828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,1,32,power_law_1.2,0.26942464828491214
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,1,32,power_law_1.2,0.2695116806030273
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,1,32,power_law_1.01,0.36159233093261717
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,1,32,power_law_1.01,1.2474419403076173
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,1,32,power_law_1.2,0.31577856063842774
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,1,32,power_law_1.2,0.39002750396728514
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,1,32,power_law_1.2,0.2797094345092773
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,1,32,power_law_1.2,0.325214729309082
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,1,32,power_law_1.01,1.531102752685547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,1,32,power_law_1.2,0.30317823410034184
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,1,32,power_law_1.2,0.36811008453369143
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,1,32,power_law_1.2,0.3491340637207031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,1,32,power_law_1.2,0.31387392044067386
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,1,32,power_law_1.2,0.3619776153564453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,1,32,power_law_1.2,0.2690496063232422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,32,1,power_law_1.2,2.326247100830078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,1,32,power_law_1.2,0.4074585723876953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,1,32,power_law_1.2,0.4095257568359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,1,32,power_law_1.2,0.4011135864257812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,1,32,power_law_1.2,0.447327995300293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,1,32,power_law_1.2,0.5359308624267578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,1,32,power_law_1.2,0.576487693786621
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,1,32,power_law_1.2,0.6975974273681641
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,1,32,power_law_1.2,0.7812339019775391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,1,32,power_law_1.2,1.1254093170166015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,1,32,power_law_1.2,0.10666496276855468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,1,32,power_law_1.2,1.3825433349609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,1,32,power_law_1.2,0.15134592056274415
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,1,32,power_law_1.2,0.15210880279541014
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,1,32,power_law_1.2,0.15399935722351074
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,1,32,power_law_1.2,0.16650239944458006
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,1,32,power_law_1.2,0.15629823684692384
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,1,32,power_law_1.2,2.3272869873046877
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,1,32,power_law_1.2,0.1289087963104248
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,1,32,power_law_1.2,0.14866559982299804
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,1,32,power_law_1.2,0.14726143836975097
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,1,32,power_law_1.2,0.17539199829101562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,1,32,power_law_1.2,2.6261273193359376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,1,32,power_law_1.2,0.17737344741821287
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,1,32,power_law_1.2,0.18131071090698242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,1,32,power_law_1.2,0.18840192794799804
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,1,32,power_law_1.2,0.17644287109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,1,32,power_law_1.2,0.1834982490539551
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,1,32,power_law_1.2,0.21960063934326174
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,1,32,power_law_1.2,0.23286144256591798
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,1,32,power_law_1.2,0.2688819122314453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,1,32,power_law_1.2,0.32672000885009767
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,1,32,power_law_1.2,0.1613875198364258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,1,32,power_law_1.2,0.38492416381835937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,1,32,power_law_1.2,0.5402777481079102
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,1,32,power_law_1.2,0.18315263748168947
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,1,32,power_law_1.2,0.7280883026123047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,2,16,balanced,0.05702400207519531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,2,16,balanced,0.05588479995727539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,2,16,balanced,0.06450943946838379
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,2,16,balanced,0.11355263710021972
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,2,16,balanced,0.18742143630981445
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,1,32,power_law_1.2,1.0182463836669922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,2,16,balanced,0.2772185516357422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,1,32,power_law_1.2,1.3798899841308594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,2,16,balanced,0.3791731262207031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,1,32,power_law_1.2,4.10521728515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,2,16,balanced,0.38472190856933597
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,2,16,balanced,0.3830323028564453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,1,32,power_law_1.2,5.46123779296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,1,32,power_law_1.2,1.8027545166015624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,2,16,balanced,0.39409278869628905
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,2,16,balanced,0.07621632099151612
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,2,16,balanced,0.3994662475585938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,2,16,balanced,0.4059763336181641
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,2,16,balanced,0.40828929901123046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,2,16,balanced,0.4296691131591797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,2,16,balanced,0.4135935974121094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,2,16,balanced,0.4537446212768555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,2,16,balanced,0.5056422424316407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,2,16,balanced,0.5536345672607422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,2,16,balanced,0.6343513488769531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,2,16,balanced,0.033314559459686276
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,2,16,balanced,0.37973121643066404
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,2,16,balanced,0.7527820587158203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,2,16,balanced,0.39746177673339844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,2,16,balanced,0.8779315185546874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,2,16,balanced,0.033251841068267826
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,2,16,balanced,0.03997184038162231
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,2,16,balanced,0.049235200881958006
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,2,16,balanced,0.0686131191253662
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,2,16,balanced,1.1878195190429688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,2,16,balanced,0.11232255935668944
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,2,16,balanced,0.18037120819091795
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,2,16,balanced,0.18121599197387694
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,2,16,balanced,0.181209602355957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,2,16,balanced,0.18766847610473633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,2,16,balanced,0.18854143142700194
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,2,16,balanced,0.18089855194091797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,2,16,balanced,0.1850214385986328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,2,16,balanced,0.18639232635498046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,2,16,balanced,0.19241344451904296
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,2,16,balanced,0.20489599227905275
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,2,16,balanced,0.22479871749877928
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,2,16,balanced,0.2495334434509277
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,2,16,balanced,0.27592575073242187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,2,16,balanced,2.1811622619628905
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,2,16,balanced,0.31993600845336917
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,2,16,balanced,0.16247039794921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,2,16,balanced,0.3688665771484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,2,16,balanced,0.18408191680908204
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,2,16,power_law_1.01,0.11598208427429199
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,2,16,balanced,1.570487060546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,2,16,power_law_1.01,0.18327552795410157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,2,16,balanced,0.6229081726074218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,2,16,power_law_1.01,0.26893312454223633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,2,16,balanced,0.8136319732666015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,2,16,power_law_1.01,0.2750540733337402
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,2,16,power_law_1.01,0.31856639862060543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,2,16,balanced,1.094823684692383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,2,16,power_law_1.01,0.30128511428833005
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,2,16,power_law_1.01,0.2963007926940918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,2,16,power_law_1.01,0.2919526481628418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,2,16,balanced,0.47031169891357416
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,2,16,power_law_1.01,0.29104383468627926
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,2,16,power_law_1.01,0.33144577026367184
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,2,16,power_law_1.01,0.30264703750610356
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,2,16,power_law_1.01,0.31063295364379884
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,2,16,power_law_1.01,0.34175743103027345
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,2,16,power_law_1.01,0.2710233688354492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,2,16,power_law_1.01,0.3512358474731445
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,2,16,power_law_1.01,0.3911948776245117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,2,16,power_law_1.01,0.36856319427490236
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,2,16,power_law_1.01,0.43512958526611334
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,2,16,power_law_1.01,0.4399769592285156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,2,16,power_law_1.01,0.578078727722168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,2,16,power_law_1.01,0.6597901153564453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,2,16,power_law_1.01,0.7903961944580078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,2,16,power_law_1.01,0.06605696201324462
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,2,16,power_law_1.01,0.9671206665039062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,2,16,power_law_1.01,0.1086092758178711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,2,16,power_law_1.01,0.1526374435424805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,2,16,power_law_1.01,0.15160320281982423
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,2,16,power_law_1.01,0.15690367698669433
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,2,16,power_law_1.01,1.662476806640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,2,16,power_law_1.01,0.15238911628723145
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,2,16,power_law_1.01,0.1689945602416992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,2,16,power_law_1.01,0.16676095962524412
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,2,16,power_law_1.01,0.17268224716186525
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,2,16,power_law_1.01,0.17514751434326173
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,2,16,power_law_1.01,2.282248992919922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,2,16,power_law_1.01,1.2501900482177735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,2,16,power_law_1.01,0.1802560043334961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,2,16,power_law_1.01,0.18179967880249023
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,2,16,power_law_1.01,0.17736320495605468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,2,16,power_law_1.01,0.17890304565429688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,2,16,power_law_1.01,0.19223295211791994
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,2,16,power_law_1.01,0.17497215270996094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,2,16,power_law_1.01,0.2063667106628418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,2,16,power_law_1.01,0.22335872650146485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,2,16,power_law_1.01,0.2556697654724121
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,2,16,power_law_1.01,0.30398719787597656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,2,16,power_law_1.01,3.1786880493164062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,2,16,power_law_1.01,0.39056896209716796
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,2,16,power_law_1.01,0.447718391418457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,2,16,power_law_1.2,0.12798848152160644
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,2,16,power_law_1.01,0.6852262115478516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,2,16,power_law_1.2,0.18403583526611328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,2,16,power_law_1.01,0.7678771209716797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,2,16,power_law_1.01,0.17936384201049804
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,2,16,power_law_1.2,0.2477350425720215
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,2,16,power_law_1.2,0.27075199127197264
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,2,16,power_law_1.2,0.27708288192749025
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,2,16,power_law_1.2,0.2924646377563477
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,2,16,power_law_1.2,0.3004006385803223
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,2,16,power_law_1.01,1.116912612915039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,2,16,power_law_1.2,0.2728025627136231
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,2,16,power_law_1.2,0.2839039993286133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,2,16,power_law_1.2,0.2967782402038574
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,2,16,power_law_1.2,0.30231552124023436
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,2,16,power_law_1.2,0.3490585708618164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,2,16,power_law_1.2,0.35681278228759766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,2,16,power_law_1.2,0.3471039962768555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,2,16,power_law_1.2,0.40246528625488287
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,2,16,power_law_1.2,0.4715276718139648
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,2,16,power_law_1.2,0.37375358581542967
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,2,16,power_law_1.2,0.4800486373901367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,2,16,power_law_1.2,0.6148492813110351
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,2,16,power_law_1.2,0.6650061035156251
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,2,16,power_law_1.01,1.513269805908203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,2,16,power_law_1.2,0.9032128143310547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,2,16,power_law_1.2,0.07165567874908448
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,2,16,power_law_1.2,1.1115122985839843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,2,16,power_law_1.2,0.3377689743041992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,2,16,power_law_1.2,0.10835968017578126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,2,16,power_law_1.2,0.14670080184936524
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,2,16,power_law_1.2,0.15349760055541992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,2,16,power_law_1.2,0.15637887954711913
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,2,16,power_law_1.2,1.629902038574219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,2,16,power_law_1.2,0.17355392456054688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,2,16,power_law_1.2,0.16205951690673828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,2,16,power_law_1.2,0.16306175231933592
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,2,16,power_law_1.2,0.17028352737426758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,2,16,power_law_1.2,0.15057663917541503
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,2,16,power_law_1.2,0.17560319900512694
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,2,16,power_law_1.2,0.18034048080444337
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,2,16,power_law_1.2,0.18475263595581054
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,2,16,power_law_1.2,0.17626495361328126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,2,16,power_law_1.2,0.17932287216186524
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,2,16,power_law_1.2,0.18148096084594728
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,2,16,power_law_1.2,0.2112998390197754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,2,16,power_law_1.2,0.2408358383178711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,2,16,power_law_1.2,0.26635007858276366
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,2,16,power_law_1.2,0.319370231628418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,2,16,power_law_1.2,2.9641741943359374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,2,16,power_law_1.2,2.0869273376464843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,2,16,power_law_1.2,0.39901824951171877
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,2,16,power_law_1.2,0.5217689514160156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,4,8,balanced,0.033518080711364744
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,2,16,power_law_1.2,0.18701183319091796
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,2,16,power_law_1.2,0.6965618896484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,4,8,balanced,0.05634431838989258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,4,8,balanced,0.08051327705383302
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,2,16,power_law_1.2,0.882956771850586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,4,8,balanced,0.11378560066223145
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,2,16,power_law_1.2,3.933952026367187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,4,8,balanced,0.2790105628967285
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,4,8,balanced,0.3659929656982422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,2,16,power_law_1.2,1.3338124084472658
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,4,8,balanced,0.3819865417480469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,4,8,balanced,0.3741555023193359
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,4,8,balanced,0.3703180694580078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,4,8,balanced,0.06470911979675294
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,4,8,balanced,0.3801497650146485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,4,8,balanced,0.38974590301513673
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,4,8,balanced,0.4046425628662109
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,2,16,power_law_1.2,1.8267007446289063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,4,8,balanced,0.18615039825439453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,4,8,balanced,0.40589569091796873
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,4,8,balanced,0.41719936370849614
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,4,8,balanced,0.42041854858398436
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,4,8,balanced,0.4485567855834961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,4,8,balanced,0.48673152923583984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,4,8,balanced,0.5546060943603516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,4,8,balanced,0.6299084854125977
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,4,8,balanced,0.033255679607391356
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,4,8,balanced,0.8848870086669922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,4,8,balanced,0.03399296045303345
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,4,8,balanced,0.04276224136352539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,4,8,balanced,0.05483136177062988
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,4,8,balanced,1.136200942993164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,4,8,balanced,0.07460991859436035
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,4,8,balanced,0.11933440208435057
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,4,8,balanced,0.1724825668334961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,4,8,balanced,1.48334716796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,4,8,balanced,0.1935116767883301
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,4,8,balanced,0.19603967666625977
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,4,8,balanced,0.1994918441772461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,4,8,balanced,0.7677171325683594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,4,8,balanced,0.19330175399780275
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,4,8,balanced,0.1954265594482422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,4,8,balanced,0.19850624084472657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,4,8,balanced,0.20427520751953124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,4,8,balanced,0.21299840927124025
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,4,8,balanced,2.041343994140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,4,8,balanced,0.2178803253173828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,4,8,balanced,0.23138944625854493
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,4,8,balanced,0.2519206428527832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,4,8,balanced,0.27733119964599606
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,4,8,balanced,0.32112640380859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,4,8,balanced,0.36361984252929686
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,4,8,balanced,0.45164543151855463
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,4,8,balanced,0.5409958267211914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,4,8,power_law_1.01,0.08089599609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,4,8,balanced,0.7171865844726563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,4,8,balanced,2.756864013671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,4,8,power_law_1.01,0.1096332836151123
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,4,8,power_law_1.01,0.17875711441040038
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,4,8,power_law_1.01,0.16221952438354492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,4,8,balanced,0.9551014709472657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,4,8,power_law_1.01,0.20444543838500975
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,4,8,power_law_1.01,0.26610944747924803
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,4,8,power_law_1.01,0.25766143798828123
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,4,8,power_law_1.01,0.32171264648437503
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,4,8,power_law_1.01,0.36779903411865233
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,4,8,power_law_1.01,0.3145523262023926
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,4,8,power_law_1.01,0.29390592575073243
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,4,8,power_law_1.01,0.2936793518066406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,4,8,power_law_1.01,0.3154508781433106
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,4,8,power_law_1.01,0.3309427261352539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,4,8,power_law_1.01,0.3648934555053711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,4,8,power_law_1.01,0.3976985549926758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,4,8,power_law_1.01,0.4707609558105469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,4,8,balanced,1.3078848266601564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,4,8,power_law_1.01,0.5555801773071289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,4,8,power_law_1.01,0.6184563064575196
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,4,8,power_law_1.01,0.7381005096435547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,4,8,balanced,1.7575462341308594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,4,8,power_law_1.01,0.055979518890380865
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,4,8,power_law_1.01,0.8818022155761719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,4,8,power_law_1.01,0.07153664112091065
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,4,8,power_law_1.01,0.3010662460327148
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,4,8,power_law_1.01,1.2443174743652343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,4,8,power_law_1.01,0.1147532844543457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,4,8,power_law_1.01,0.10549504280090331
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,4,8,power_law_1.01,0.32003456115722656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,4,8,power_law_1.01,1.507439422607422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,4,8,power_law_1.01,0.14098943710327147
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,4,8,power_law_1.01,0.14278016090393067
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,4,8,power_law_1.01,0.15997695922851562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,4,8,power_law_1.01,0.1863257598876953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,4,8,power_law_1.01,0.17640703201293945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,4,8,power_law_1.01,0.1893132781982422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,4,8,power_law_1.01,0.18942975997924805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,4,8,power_law_1.01,0.18476032257080077
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,4,8,power_law_1.01,0.18710912704467772
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,4,8,power_law_1.01,2.2544691467285154
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,4,8,power_law_1.01,0.1939084815979004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,4,8,power_law_1.01,0.19887487411499022
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,4,8,power_law_1.01,0.20598272323608396
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,4,8,power_law_1.01,0.23349119186401368
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,4,8,power_law_1.01,0.24783359527587892
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,4,8,power_law_1.01,0.32420352935791014
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,4,8,power_law_1.01,0.43870208740234373
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,4,8,power_law_1.01,0.1778099250793457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,4,8,power_law_1.01,2.9420660400390624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,4,8,power_law_1.01,0.5211532974243165
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,4,8,power_law_1.01,0.7056972503662109
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,4,8,power_law_1.2,0.09065600395202636
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,4,8,power_law_1.2,0.10959744453430176
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,4,8,power_law_1.2,0.13865856170654295
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,4,8,power_law_1.2,0.16411008834838867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,4,8,power_law_1.2,0.1988479995727539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,4,8,power_law_1.01,0.2841292762756348
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,4,8,power_law_1.01,1.301651153564453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,4,8,power_law_1.2,0.26785024642944333
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,4,8,power_law_1.2,0.355918083190918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,4,8,power_law_1.2,0.2908198356628418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,4,8,power_law_1.01,1.6587852478027343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,4,8,power_law_1.2,0.29186943054199216
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,4,8,power_law_1.2,0.2719615936279297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,4,8,power_law_1.01,0.8935270690917969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,4,8,power_law_1.2,0.2786816024780273
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,4,8,power_law_1.2,0.2857561683654785
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,4,8,power_law_1.2,0.3172518348693848
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,4,8,power_law_1.2,0.31453439712524417
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,4,8,power_law_1.2,0.35779457092285155
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,4,8,power_law_1.2,0.2557964706420898
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,4,8,power_law_1.2,0.3535820770263672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,4,8,power_law_1.2,0.4303283309936523
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,4,8,power_law_1.2,0.46130943298339844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,4,8,power_law_1.2,0.5536140823364258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,4,8,power_law_1.2,0.655546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,4,8,power_law_1.2,0.7998758697509766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,4,8,power_law_1.2,0.0521395206451416
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,4,8,power_law_1.2,0.07192063808441163
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,4,8,power_law_1.2,0.958023681640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,4,8,power_law_1.2,0.0865817642211914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,4,8,power_law_1.2,0.10537343978881837
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,4,8,power_law_1.2,0.1373516845703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,4,8,power_law_1.2,1.390830078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,4,8,power_law_1.2,0.16018815994262697
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,4,8,power_law_1.2,0.15307264328002929
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,4,8,power_law_1.2,0.1887526321411133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,4,8,power_law_1.2,1.6311386108398438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,4,8,power_law_1.2,0.17451904296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,4,8,power_law_1.2,0.15270015716552734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,4,8,power_law_1.2,0.1915443229675293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,4,8,power_law_1.2,0.18717311859130858
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,4,8,power_law_1.2,0.1839027214050293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,4,8,power_law_1.2,0.18571136474609376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,4,8,power_law_1.2,0.193438720703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,4,8,power_law_1.2,0.1979840087890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,4,8,power_law_1.2,0.20743295669555667
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,4,8,power_law_1.2,0.233493766784668
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,4,8,power_law_1.2,0.25396480560302737
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,4,8,power_law_1.2,2.519370269775391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,4,8,power_law_1.2,0.29413375854492185
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,4,8,power_law_1.2,0.3520025634765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,8,4,balanced,0.03125119924545288
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,4,8,power_law_1.2,0.5544063949584961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,8,4,balanced,0.05528448104858399
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,8,4,balanced,0.06411136150360107
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,4,8,power_law_1.2,0.7621580505371094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,8,4,balanced,0.08251520156860351
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,4,8,power_law_1.2,3.2056564331054687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,8,4,balanced,0.11684736251831054
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,4,8,power_law_1.2,0.9667340850830077
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,8,4,balanced,0.18671104431152344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,8,4,balanced,0.274330883026123
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,8,4,balanced,0.27928319931030277
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,8,4,balanced,0.2919398307800293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,8,4,balanced,0.28233856201171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,4,8,power_law_1.2,1.4221235656738282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,8,4,balanced,0.3636095809936523
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,8,4,balanced,0.36982017517089844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,8,4,balanced,0.37384193420410156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,8,4,balanced,0.3806643295288086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,4,8,power_law_1.2,1.8695269775390624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,8,4,balanced,0.40736129760742185
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,8,4,balanced,0.3912960052490234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,8,4,balanced,0.44042110443115234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,8,4,balanced,0.47325439453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,8,4,balanced,0.5111833572387695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,4,8,power_law_1.2,0.48149120330810546
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,8,4,balanced,0.6035302352905274
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,8,4,balanced,0.7290278625488281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,8,4,balanced,0.03125888109207153
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,8,4,balanced,0.9135616302490235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,8,4,balanced,0.033219840526580814
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,8,4,balanced,0.04286719799041748
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,8,4,balanced,0.05708799839019776
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,8,4,balanced,0.08423423767089844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,8,4,balanced,1.154375686645508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,8,4,balanced,0.1321331214904785
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,8,4,balanced,0.19003007888793946
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,8,4,balanced,0.19061376571655272
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,8,4,balanced,0.1934553527832031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,8,4,balanced,0.19446399688720703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,8,4,balanced,0.36365055084228515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,8,4,balanced,0.2337753677368164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,8,4,balanced,0.23565439224243162
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,8,4,balanced,0.2397273635864258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,8,4,balanced,0.24328575134277344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,8,4,balanced,0.24985855102539062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,8,4,balanced,0.2535807991027832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,8,4,balanced,0.26220800399780275
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,8,4,balanced,2.046771240234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,8,4,balanced,0.28498048782348634
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,8,4,balanced,0.3045414352416992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,8,4,balanced,0.34663169860839843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,8,4,balanced,0.3848729705810547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,8,4,balanced,0.4693286514282226
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,8,4,balanced,0.5532032012939453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,8,4,balanced,2.7662234497070313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,8,4,balanced,0.722007064819336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,8,4,power_law_1.01,0.06402560234069823
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,8,4,power_law_1.01,0.08488320350646973
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,8,4,balanced,0.9964006042480469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,8,4,power_law_1.01,0.10724991798400878
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,8,4,power_law_1.01,0.10847616195678711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,8,4,power_law_1.01,0.14534655570983887
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,8,4,balanced,1.4902784729003906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,8,4,power_law_1.01,0.17774208068847658
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,8,4,power_law_1.01,0.24153087615966795
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,8,4,power_law_1.01,0.21317247390747068
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,8,4,power_law_1.01,0.23102720260620116
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,8,4,power_law_1.01,0.2823923110961914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,8,4,power_law_1.01,0.29705984115600587
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,8,4,balanced,1.8017677307128905
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,8,4,power_law_1.01,0.2831769561767578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,8,4,power_law_1.01,0.30332544326782224
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,8,4,power_law_1.01,0.3102937507629394
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,8,4,power_law_1.01,0.3084147262573242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,8,4,balanced,1.3351834106445313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,8,4,power_law_1.01,0.32529281616210937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,8,4,power_law_1.01,0.3841727828979492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,8,4,power_law_1.01,0.4288998413085937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,8,4,power_law_1.01,0.5277030563354492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,8,4,power_law_1.01,0.6563852691650391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,8,4,power_law_1.01,0.7740338897705078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,8,4,power_law_1.01,0.04113024234771728
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,8,4,power_law_1.01,0.17298816680908202
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,8,4,power_law_1.01,0.05474559783935547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,8,4,power_law_1.01,1.2089881896972656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,8,4,power_law_1.01,0.07643519878387452
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,8,4,power_law_1.01,0.07678592205047607
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,8,4,power_law_1.01,0.1090611171722412
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,8,4,power_law_1.01,1.5414643859863282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,8,4,power_law_1.01,0.1354751968383789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,8,4,power_law_1.01,0.1774348831176758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,8,4,power_law_1.01,0.16128768920898437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,8,4,power_law_1.01,0.9008767700195313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,8,4,power_law_1.01,0.16776063919067383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,8,4,power_law_1.01,0.20580480575561522
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,8,4,power_law_1.01,2.1726002502441406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,8,4,power_law_1.01,0.2154739189147949
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,8,4,power_law_1.01,0.21445760726928711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,8,4,power_law_1.01,0.2235737609863281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,8,4,power_law_1.01,0.22977279663085937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,8,4,power_law_1.01,0.2342310333251953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,8,4,power_law_1.01,0.2470899200439453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,8,4,power_law_1.01,0.28452224731445314
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,8,4,power_law_1.01,0.129039363861084
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,8,4,power_law_1.01,0.3341081619262695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,8,4,power_law_1.01,2.83093505859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,8,4,power_law_1.01,0.37440513610839843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,8,4,power_law_1.01,0.4781644821166992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,8,4,power_law_1.2,0.06436992168426514
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,8,4,power_law_1.01,0.6089612960815429
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,8,4,power_law_1.2,0.07666304111480712
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,8,4,power_law_1.2,0.08956031799316407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,8,4,power_law_1.01,0.7950131225585937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,8,4,power_law_1.2,0.11200896263122559
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,8,4,power_law_1.2,0.1392294406890869
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,8,4,power_law_1.01,0.9996031951904296
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,8,4,power_law_1.2,0.16966272354125977
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,8,4,power_law_1.2,0.1712396812438965
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,8,4,power_law_1.2,0.2404287910461426
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,8,4,power_law_1.2,0.22284543991088865
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,8,4,power_law_1.2,0.2212428855895996
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,8,4,power_law_1.01,1.4195404052734375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,8,4,power_law_1.2,0.2786956787109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,8,4,power_law_1.2,0.2752716827392578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,8,4,power_law_1.2,0.293120002746582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,8,4,power_law_1.2,0.3005695915222168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,8,4,power_law_1.2,0.3138431930541992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,8,4,power_law_1.2,0.2965785598754883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,8,4,power_law_1.01,0.2665395164489746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,8,4,power_law_1.2,0.3381990432739258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,8,4,power_law_1.2,0.41506305694580076
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,8,4,power_law_1.2,0.454334716796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,8,4,power_law_1.2,0.577916145324707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,8,4,power_law_1.2,0.6722188568115234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,8,4,power_law_1.2,0.7823296356201171
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,8,4,power_law_1.2,0.0409881591796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,8,4,power_law_1.2,0.962823715209961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,8,4,power_law_1.01,1.8346829223632812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,8,4,power_law_1.2,0.054260478019714356
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,8,4,power_law_1.2,0.06520448207855226
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,8,4,power_law_1.2,1.2839360046386719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,8,4,power_law_1.2,0.07785855770111084
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,8,4,power_law_1.2,0.10933759689331055
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,8,4,power_law_1.2,0.13214336395263673
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,8,4,power_law_1.2,1.6065100097656249
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,8,4,power_law_1.2,0.12805248260498048
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,8,4,power_law_1.2,0.15652095794677734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,8,4,power_law_1.2,0.1693619155883789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,8,4,power_law_1.2,0.15772928237915038
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,8,4,power_law_1.2,0.2035532760620117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,8,4,power_law_1.2,0.21092735290527345
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,8,4,power_law_1.2,0.2220403289794922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,8,4,power_law_1.2,0.2285081672668457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,8,4,power_law_1.2,0.22733951568603517
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,8,4,power_law_1.2,0.23996671676635742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,8,4,power_law_1.2,0.2686425590515137
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,8,4,power_law_1.2,0.289483528137207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,8,4,power_law_1.2,0.34237438201904297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,8,4,power_law_1.2,0.3907340621948242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,8,4,power_law_1.2,0.5118592071533203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,8,4,power_law_1.2,3.0818136596679686
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,8,4,power_law_1.2,0.21737087249755857
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,8,4,power_law_1.2,0.6311808013916016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,8,4,power_law_1.2,2.3333171081542967
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,16,2,balanced,0.056062722206115724
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,8,4,power_law_1.2,0.8476914978027343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,16,2,balanced,0.0704089593887329
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,16,2,balanced,0.08221183776855469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,16,2,balanced,0.12193663597106934
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,8,4,power_law_1.2,1.029345245361328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,16,2,balanced,0.2051571273803711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,16,2,balanced,0.2912076759338379
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,16,2,balanced,0.29419008255004886
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,16,2,balanced,0.29793535232543944
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,16,2,balanced,0.29617279052734374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,16,2,balanced,0.0340339207649231
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,16,2,balanced,0.30402687072753903
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,8,4,power_law_1.2,1.5086373901367187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,16,2,balanced,0.3088678359985352
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,16,2,balanced,0.38145790100097654
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,16,2,balanced,0.3873497772216797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,16,2,balanced,0.39140350341796876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,16,2,balanced,0.4004582214355469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,8,4,power_law_1.2,1.9439117431640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,16,2,balanced,0.44553470611572266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,16,2,balanced,0.5321484756469727
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,16,2,balanced,0.6005670547485351
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,16,2,balanced,0.7373324584960937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,16,2,balanced,0.9243596649169922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,16,2,balanced,0.31487615585327144
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,16,2,balanced,0.03564287900924683
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,16,2,balanced,0.046856961250305175
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,16,2,balanced,1.216855010986328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,16,2,balanced,0.06554880142211914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,16,2,balanced,0.4235673522949218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,16,2,balanced,0.10040191650390624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,16,2,balanced,0.17402368545532226
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,16,2,balanced,0.2446873664855957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,16,2,balanced,0.24653568267822265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,16,2,balanced,0.24941823959350584
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,16,2,balanced,0.25019775390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,16,2,balanced,0.03184639930725098
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,16,2,balanced,0.2526348876953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,16,2,balanced,0.2570841598510742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,16,2,balanced,0.26039167404174807
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,16,2,balanced,0.32944000244140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,16,2,balanced,0.33563262939453126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,16,2,balanced,0.33595905303955076
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,16,2,balanced,0.34444030761718747
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,16,2,balanced,0.36246910095214846
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,16,2,balanced,1.5867301940917968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,16,2,balanced,0.41662719726562497
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,16,2,balanced,0.45346817016601565
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,16,2,balanced,0.5333887863159179
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,16,2,balanced,0.610200309753418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,16,2,power_law_1.01,0.05483263969421387
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,16,2,balanced,2.1950950622558594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,16,2,balanced,0.7659020996093749
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,16,2,power_law_1.01,0.06455039978027344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,16,2,power_law_1.01,0.07918464183807374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,16,2,balanced,0.37820545196533206
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,16,2,power_law_1.01,0.09533056259155273
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,16,2,power_law_1.01,0.12644224166870116
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,16,2,balanced,2.975658264160156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,16,2,power_law_1.01,0.16544511795043945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,16,2,balanced,1.4356288146972656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,16,2,power_law_1.01,0.2231180763244629
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,16,2,power_law_1.01,0.22642047882080077
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,16,2,power_law_1.01,0.23758720397949218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,16,2,power_law_1.01,0.23625215530395507
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,16,2,power_law_1.01,0.26579072952270505
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,16,2,balanced,1.957840576171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,16,2,power_law_1.01,0.301529598236084
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,16,2,balanced,1.1114854431152343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,16,2,power_law_1.01,0.30271615982055666
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,16,2,power_law_1.01,0.30816255569458006
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,16,2,power_law_1.01,0.3341158294677734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,16,2,power_law_1.01,0.18374656677246093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,16,2,power_law_1.01,0.3771366500854492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,16,2,power_law_1.01,0.42624641418457027
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,16,2,power_law_1.01,0.5618726348876952
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,16,2,power_law_1.01,0.620057601928711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,16,2,power_law_1.01,0.2543398475646973
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,16,2,power_law_1.01,0.7728486633300781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,16,2,power_law_1.01,0.03519232034683227
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,16,2,power_law_1.01,0.9097676849365234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,16,2,power_law_1.01,0.046949119567871095
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,16,2,power_law_1.01,0.06295936107635498
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,16,2,power_law_1.01,0.07632383823394776
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,16,2,power_law_1.01,0.10741375923156737
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,16,2,power_law_1.01,0.1341273593902588
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,16,2,power_law_1.01,0.15265664100646972
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,16,2,power_law_1.01,1.5759577941894531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,16,2,power_law_1.01,0.1988070487976074
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,16,2,power_law_1.01,0.20412672042846677
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,16,2,power_law_1.01,0.21046911239624025
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,16,2,power_law_1.01,2.1898226928710938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,16,2,power_law_1.01,0.2225062370300293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,16,2,power_law_1.01,0.2234649658203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,16,2,power_law_1.01,1.271234588623047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,16,2,power_law_1.01,0.29250431060791016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,16,2,power_law_1.01,0.2955046463012695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,16,2,power_law_1.01,0.3091967964172363
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,16,2,power_law_1.01,0.33488639831542966
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,16,2,power_law_1.01,0.3530060958862305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,16,2,power_law_1.01,2.82468994140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,16,2,power_law_1.01,0.2140070343017578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,16,2,power_law_1.01,0.41641342163085937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,16,2,power_law_1.01,0.4602099227905273
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,16,2,power_law_1.01,0.5705420684814453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,16,2,power_law_1.01,0.307458553314209
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,16,2,power_law_1.2,0.05558784008026123
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,16,2,power_law_1.01,0.9202098846435547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,16,2,power_law_1.2,0.06429952144622803
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,16,2,power_law_1.2,0.07432960033416748
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,16,2,power_law_1.2,0.09243136405944824
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,16,2,power_law_1.01,1.1592320251464845
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,16,2,power_law_1.2,0.12152576446533203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,16,2,power_law_1.2,0.15278847694396974
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,16,2,power_law_1.2,0.1737548828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,16,2,power_law_1.2,0.22163455963134765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,16,2,power_law_1.01,1.5656665039062498
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,16,2,power_law_1.2,0.22960895538330078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,16,2,power_law_1.01,0.703342056274414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,16,2,power_law_1.2,0.24709888458251955
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,16,2,power_law_1.2,0.27778175354003903
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,16,2,power_law_1.2,0.29865983963012693
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,16,2,power_law_1.2,0.307010555267334
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,16,2,power_law_1.01,2.0420314025878907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,16,2,power_law_1.2,0.31091583251953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,16,2,power_law_1.2,0.3440755081176758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,16,2,power_law_1.2,0.39091712951660157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,16,2,power_law_1.2,0.4307020950317383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,16,2,power_law_1.2,0.2269580841064453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,16,2,power_law_1.2,0.23921024322509768
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,16,2,power_law_1.2,0.6864409637451172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,16,2,power_law_1.2,0.781928939819336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,16,2,power_law_1.2,0.929164810180664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,16,2,power_law_1.2,0.03507200002670288
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,16,2,power_law_1.2,0.04725759983062744
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,16,2,power_law_1.2,0.05805696010589599
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,16,2,power_law_1.2,0.07687039852142334
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,16,2,power_law_1.2,0.10359807968139649
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,16,2,power_law_1.2,1.3106597900390624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,16,2,power_law_1.2,0.5519961547851563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,16,2,power_law_1.2,0.14310912132263182
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,16,2,power_law_1.2,1.6121778869628904
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,16,2,power_law_1.2,0.1926976013183594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,16,2,power_law_1.2,0.2012928009033203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,16,2,power_law_1.2,0.2061235237121582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,16,2,power_law_1.2,0.217706241607666
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,16,2,power_law_1.2,0.216888313293457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,16,2,power_law_1.2,0.2254707145690918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,16,2,power_law_1.2,0.29091199874877927
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,16,2,power_law_1.2,0.29823232650756837
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,16,2,power_law_1.2,2.24395263671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,16,2,power_law_1.2,0.2972851181030273
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,16,2,power_law_1.2,0.3057561683654785
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,16,2,power_law_1.2,0.33583999633789063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,16,2,power_law_1.2,0.13171072006225587
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,16,2,power_law_1.2,0.3586649703979492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,16,2,power_law_1.2,0.42073726654052734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,16,2,power_law_1.2,0.48425727844238275
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,16,2,power_law_1.2,2.873077697753906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,16,2,power_law_1.2,0.5943167877197266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,32,1,balanced,0.05425151824951172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,32,1,balanced,0.0710259199142456
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,32,1,balanced,0.08102975845336914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,16,2,power_law_1.2,0.7179366302490234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,32,1,balanced,0.10269568443298341
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,16,2,power_law_1.2,0.9303103637695311
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,32,1,balanced,0.14920767784118652
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,32,1,balanced,0.250316162109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,16,2,power_law_1.2,1.1738304138183593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,32,1,balanced,0.36980480194091797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,32,1,balanced,0.37239166259765627
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,32,1,balanced,0.3764326477050781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,32,1,balanced,0.37238719940185544
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,32,1,balanced,0.37959167480468753
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,16,2,power_law_1.2,1.6098355102539064
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,32,1,balanced,0.39022464752197267
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,32,1,balanced,0.4001068878173828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,32,1,balanced,0.40920318603515626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,32,1,balanced,0.4148831939697265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,32,1,balanced,0.45029312133789057
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,16,2,power_law_1.2,2.0969215393066407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,32,1,balanced,0.47396480560302734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,32,1,balanced,0.5807360076904298
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,32,1,balanced,0.6490425872802734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,32,1,balanced,0.38418815612792967
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,32,1,balanced,0.7959020996093751
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,32,1,balanced,0.0448089599609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,32,1,balanced,0.5005158233642578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,32,1,balanced,0.05250048160552978
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,32,1,balanced,1.1695155334472656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,32,1,balanced,0.06898047924041747
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,32,1,balanced,0.10019968032836915
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,32,1,balanced,1.2425791931152343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,32,1,balanced,0.1532863998413086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,32,1,balanced,0.26808576583862304
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,32,1,balanced,0.3880780792236328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,32,1,balanced,0.3898291015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,32,1,balanced,0.392279052734375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,32,1,balanced,1.7409695434570314
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,32,1,balanced,0.39501247406005857
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,32,1,balanced,0.3959462356567383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,32,1,balanced,0.4002848052978516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,32,1,balanced,0.40238208770751954
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,32,1,balanced,0.4085977554321289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,32,1,balanced,0.41519039154052734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,32,1,balanced,0.4187468719482422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,32,1,balanced,2.473157196044922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,32,1,balanced,0.5558233642578125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,32,1,balanced,0.5742745590209961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,32,1,balanced,0.5887238311767578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,32,1,balanced,0.6270796966552734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,32,1,balanced,0.6635129547119141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,32,1,balanced,0.7452448272705079
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,32,1,power_law_1.01,0.04400959968566894
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,32,1,balanced,0.8383020782470704
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,32,1,power_law_1.01,0.06991744041442871
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,32,1,balanced,0.9915372467041015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,32,1,balanced,3.601944274902344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,32,1,power_law_1.01,0.0809721565246582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,32,1,power_law_1.01,0.10243904113769531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,32,1,power_law_1.01,0.1452294445037842
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,32,1,power_law_1.01,0.1968550491333008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,32,1,power_law_1.01,0.2308652877807617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,32,1,balanced,1.5629017639160157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,32,1,power_law_1.01,0.2774131202697754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,32,1,power_law_1.01,0.2928416061401367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,32,1,power_law_1.01,0.3046380805969238
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,32,1,balanced,1.899333801269531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,32,1,power_law_1.01,0.31812864303588867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,32,1,power_law_1.01,0.32625022888183597
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,32,1,power_law_1.01,0.34193279266357424
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,32,1,power_law_1.01,0.35860927581787105
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,32,1,power_law_1.01,0.3778355026245117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,32,1,power_law_1.01,0.41668094635009767
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,32,1,power_law_1.01,0.3701567840576172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,32,1,power_law_1.01,0.4199398422241211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,32,1,power_law_1.01,0.46734272003173827
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,32,1,balanced,2.6725619506835936
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,32,1,power_law_1.01,0.5729900741577149
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,32,1,power_law_1.01,0.6632012939453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,32,1,power_law_1.01,0.04523263931274414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,32,1,power_law_1.01,0.8591584014892579
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,32,1,power_law_1.01,0.05192448139190674
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,32,1,power_law_1.01,0.976616973876953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,32,1,power_law_1.01,0.06852287769317626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,32,1,power_law_1.01,0.09819007873535156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,32,1,power_law_1.01,0.14445247650146484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,32,1,power_law_1.01,1.3519212341308593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,32,1,power_law_1.01,0.20142400741577146
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,32,1,power_law_1.01,0.227587833404541
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,32,1,power_law_1.01,0.27693311691284184
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,32,1,power_law_1.01,1.6921279907226563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,32,1,power_law_1.01,0.29558719635009767
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,32,1,power_law_1.01,0.31574655532836915
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,32,1,power_law_1.01,0.32669185638427733
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,32,1,power_law_1.01,0.3316569519042969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,32,1,power_law_1.01,0.34096126556396483
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,32,1,power_law_1.01,0.35763648986816404
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,32,1,power_law_1.01,2.442535095214844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,32,1,power_law_1.01,0.3673235321044922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,32,1,power_law_1.01,0.49208961486816405
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,32,1,power_law_1.01,0.5195270538330078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,32,1,power_law_1.01,0.5484703826904297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,32,1,power_law_1.01,0.6139744186401368
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,32,1,power_law_1.01,0.35193023681640623
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,32,1,power_law_1.01,3.1484161376953126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,32,1,power_law_1.01,0.6929747009277344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,32,1,power_law_1.01,0.8458483123779297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,32,1,power_law_1.2,0.0439628791809082
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,32,1,power_law_1.2,0.06928127765655517
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,32,1,power_law_1.01,0.9762169647216797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,32,1,power_law_1.2,0.08122112274169921
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,32,1,power_law_1.2,0.10250240325927736
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,32,1,power_law_1.01,1.2834214782714843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,32,1,power_law_1.2,0.19205951690673828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,32,1,power_law_1.2,0.21972288131713866
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,32,1,power_law_1.2,0.2719481658935547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,32,1,power_law_1.01,1.5735385131835937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,32,1,power_law_1.2,0.2844876861572266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,32,1,power_law_1.2,0.29193920135498047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,32,1,power_law_1.2,0.30647743225097657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,32,1,power_law_1.2,0.13606911659240722
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,32,1,power_law_1.2,0.32630462646484376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,32,1,power_law_1.2,0.33952640533447265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,32,1,power_law_1.2,0.3589894485473633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,32,1,power_law_1.01,2.182469787597656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,32,1,power_law_1.2,0.37606590270996093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,32,1,power_law_1.2,0.3766035079956055
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,32,1,power_law_1.2,0.4174700927734375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,32,1,power_law_1.2,0.43057472229003907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,32,1,power_law_1.2,0.47953151702880864
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,32,1,power_law_1.01,2.8005709838867188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,32,1,power_law_1.2,0.5929427337646485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,32,1,power_law_1.2,0.716223373413086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,32,1,power_law_1.2,0.045527682304382325
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,32,1,power_law_1.2,0.8766105651855469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,32,1,power_law_1.2,1.0027820587158203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,32,1,power_law_1.2,0.05186431884765626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,32,1,power_law_1.2,0.06905920028686523
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,32,1,power_law_1.2,0.0982476806640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,32,1,power_law_1.2,0.1354271984100342
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,32,1,power_law_1.2,0.18306751251220704
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,32,1,power_law_1.2,1.7154547119140624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,32,1,power_law_1.2,0.27272192001342777
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,32,1,power_law_1.2,0.2867084884643555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,32,1,power_law_1.2,1.3645567321777343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,32,1,power_law_1.2,0.29960384368896487
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,32,1,power_law_1.2,0.3127027130126953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,32,1,power_law_1.2,0.3287411117553711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,32,1,power_law_1.2,2.449890594482422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,32,1,power_law_1.2,0.21322816848754883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,32,1,power_law_1.2,0.33883262634277345
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,32,1,power_law_1.2,0.3479667282104492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,32,1,power_law_1.2,0.35748992919921874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,32,1,power_law_1.2,0.36519935607910153
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,32,1,power_law_1.2,0.49454273223876954
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,32,1,power_law_1.2,3.1708108520507814
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,32,1,power_law_1.2,0.5302668762207031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,32,1,power_law_1.2,0.5615763092041016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,32,1,power_law_1.2,0.6421228790283203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,1,32,balanced,0.02371840000152588
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,32,1,power_law_1.2,0.7352384185791015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,1,32,balanced,0.023527679443359376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,1,32,balanced,0.02358144044876099
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,32,1,power_law_1.2,0.8787251281738282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,1,32,balanced,0.04139776229858398
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,1,32,balanced,0.05006591796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,32,1,power_law_1.2,1.0402137756347656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,1,32,balanced,0.07055744171142578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,1,32,balanced,0.09930624008178711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,1,32,balanced,0.0992460823059082
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,1,32,balanced,0.10034175872802735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,1,32,balanced,0.10139519691467284
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,1,32,balanced,0.10484479904174804
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,1,32,balanced,0.10779007911682129
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,1,32,balanced,0.10786944389343263
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,1,32,balanced,0.11127296447753907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,1,32,balanced,0.11563903808593749
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,1,32,balanced,0.10859007835388183
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,1,32,balanced,0.11531264305114745
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,1,32,balanced,0.12204031944274903
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,1,32,balanced,0.137639684677124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,1,32,balanced,0.14117504119873048
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,1,32,balanced,0.18855039596557616
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,1,32,balanced,0.21650304794311523
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,32,1,power_law_1.2,1.6500128173828124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,1,32,balanced,0.2953740882873535
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,1,32,balanced,0.10090239524841309
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,1,32,balanced,0.3402739334106445
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,1,32,balanced,0.02547584056854248
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,1,32,balanced,0.025826559066772458
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,1,32,balanced,0.025103359222412108
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,1,32,balanced,0.025975039005279543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,1,32,balanced,0.03555072069168091
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,1,32,balanced,0.048830718994140626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,32,1,power_law_1.2,2.246793670654297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,1,32,balanced,0.05319168090820312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,1,32,balanced,0.054400000572204586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,1,32,balanced,0.05545728206634522
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,1,32,balanced,0.05636864185333253
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,1,32,balanced,0.05772799968719482
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,1,32,balanced,0.059182081222534176
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,1,32,balanced,0.06409215927124023
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,1,32,balanced,0.06749184131622314
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,1,32,balanced,0.07058688163757323
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,1,32,balanced,0.06268671989440919
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,1,32,balanced,0.0695692777633667
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,1,32,balanced,0.07346432209014893
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,32,1,power_law_1.2,2.8698220825195313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,1,32,balanced,0.5009600067138672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,1,32,balanced,0.08324480056762695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,1,32,balanced,0.08997247695922853
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,1,32,balanced,0.10947711944580078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,1,32,balanced,0.6275942230224609
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,1,32,balanced,0.05348608016967773
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,1,32,balanced,0.17538047790527345
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,1,32,balanced,0.20906240463256837
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,1,32,balanced,0.29970304489135746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,1,32,power_law_1.01,0.06942336082458496
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,1,32,power_law_1.01,0.06966911792755127
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,1,32,power_law_1.01,0.06967296123504638
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,1,32,balanced,0.3641881561279297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,1,32,power_law_1.01,0.07007999897003174
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,1,32,power_law_1.01,0.09672191619873047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,1,32,power_law_1.01,0.09073535919189453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,1,32,power_law_1.01,0.09768447875976563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,1,32,power_law_1.01,0.0991539192199707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,1,32,power_law_1.01,0.09979647636413574
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,1,32,power_law_1.01,0.10395392417907715
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,1,32,power_law_1.01,0.10472703933715821
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,1,32,power_law_1.01,0.10749055862426757
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,1,32,power_law_1.01,0.11132927894592286
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,1,32,power_law_1.01,0.11586560249328612
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,1,32,power_law_1.01,0.11079296112060547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,1,32,power_law_1.01,0.11683456420898437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,32,1,power_law_1.2,1.3426911926269531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,1,32,balanced,0.12896896362304688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,1,32,power_law_1.01,0.14496000289916994
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,1,32,power_law_1.01,0.17197824478149415
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,1,32,power_law_1.01,0.22282239913940433
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,1,32,power_law_1.01,0.27571327209472657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,1,32,power_law_1.01,0.31091968536376957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,1,32,power_law_1.01,0.45449470520019536
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,1,32,power_law_1.01,0.047548160552978516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,1,32,power_law_1.01,0.5772825622558594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,1,32,power_law_1.01,0.7139750671386719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,1,32,power_law_1.01,0.04764671802520752
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,1,32,power_law_1.01,0.049030399322509764
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,1,32,power_law_1.01,0.04925183773040771
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,1,32,power_law_1.01,0.12318592071533203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,1,32,power_law_1.01,0.0508838415145874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,1,32,power_law_1.01,1.037817611694336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,1,32,power_law_1.01,0.0485427188873291
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,1,32,power_law_1.01,0.052779521942138675
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,1,32,power_law_1.01,0.05501696109771729
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,1,32,power_law_1.01,0.056884479522705075
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,1,32,power_law_1.01,0.05911295890808106
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,1,32,power_law_1.01,0.06394239902496338
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,1,32,power_law_1.01,1.3153701782226563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,1,32,power_law_1.01,0.06616064071655273
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,1,32,power_law_1.01,0.0656217622756958
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,1,32,power_law_1.01,0.07073791980743407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,1,32,power_law_1.01,0.06626688003540039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,1,32,power_law_1.01,0.0700928020477295
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,1,32,power_law_1.01,0.09452672004699707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,1,32,power_law_1.01,0.12539775848388673
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,1,32,power_law_1.01,0.1418995189666748
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,1,32,power_law_1.01,0.18363008499145508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,1,32,power_law_1.01,0.24148351669311524
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,1,32,power_law_1.01,0.3060044860839844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,1,32,power_law_1.2,0.06931968212127686
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,1,32,power_law_1.01,0.0533516788482666
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,1,32,power_law_1.2,0.06950143814086915
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,1,32,power_law_1.2,0.06987904071807861
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,1,32,power_law_1.2,0.0936575984954834
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,1,32,power_law_1.01,0.7973452758789062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,1,32,power_law_1.01,0.08210432052612304
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,1,32,power_law_1.2,0.09843839645385742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,1,32,power_law_1.2,0.10226431846618653
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,1,32,power_law_1.2,0.10623488426208497
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,1,32,power_law_1.2,0.10988927841186524
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,1,32,power_law_1.01,0.43308673858642577
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,1,32,power_law_1.2,0.1121561622619629
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,1,32,power_law_1.2,0.06951424121856689
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,1,32,power_law_1.2,0.11388031959533691
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,1,32,power_law_1.2,0.11844223976135253
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,1,32,power_law_1.2,0.09283328056335449
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,1,32,power_law_1.2,0.12839296340942383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,1,32,power_law_1.2,0.09817855834960938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,1,32,power_law_1.2,0.15608448028564453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,1,32,power_law_1.2,0.1785087966918945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,1,32,power_law_1.2,0.22982143402099608
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,1,32,power_law_1.2,0.10153471946716308
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,1,32,power_law_1.2,0.2962879943847656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,1,32,power_law_1.2,0.3723673629760742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,1,32,power_law_1.2,0.5076876831054687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,1,32,power_law_1.2,0.04709504127502441
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,1,32,power_law_1.01,0.5581427383422851
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,1,32,power_law_1.2,0.0479475212097168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,1,32,power_law_1.2,0.7285504150390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,1,32,power_law_1.2,0.11627391815185546
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,1,32,power_law_1.2,0.0485478401184082
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,1,32,power_law_1.2,0.04796160221099853
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,1,32,power_law_1.2,0.05013887882232666
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,1,32,power_law_1.2,0.0532044792175293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,1,32,power_law_1.2,0.050807042121887205
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,1,32,power_law_1.2,1.0972582244873048
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,1,32,power_law_1.2,0.053582081794738765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,1,32,power_law_1.2,0.056442880630493165
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,1,32,power_law_1.2,0.05707007884979248
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,1,32,power_law_1.2,0.0610316801071167
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,1,32,power_law_1.2,0.06395008087158202
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,1,32,power_law_1.2,0.06805888175964356
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,1,32,power_law_1.2,0.06704639911651611
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,1,32,power_law_1.2,0.06708352088928222
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,1,32,power_law_1.2,0.07672832012176514
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,1,32,power_law_1.2,1.5415769958496095
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,1,32,power_law_1.2,0.08954239845275878
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,1,32,power_law_1.2,0.10882559776306153
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,1,32,power_law_1.2,0.1349350357055664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,1,32,power_law_1.2,0.17340160369873048
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,1,32,power_law_1.2,0.21511680603027344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,1,32,power_law_1.2,0.27806079864501954
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,1,32,power_law_1.2,0.4378905487060547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,2,16,balanced,0.02158207893371582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,1,32,power_law_1.2,0.5365311813354492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,2,16,balanced,0.022958080768585205
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,1,32,power_law_1.2,0.06696191787719727
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,2,16,balanced,0.03937279939651489
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,2,16,balanced,0.049708800315856935
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,2,16,balanced,0.07182720184326172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,2,16,balanced,0.09964799880981445
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,2,16,balanced,0.09760000228881835
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,2,16,balanced,0.10106368064880371
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,2,16,balanced,0.10075648307800293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,2,16,balanced,0.10330495834350586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,2,16,balanced,0.10547327995300293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,2,16,balanced,0.11048831939697265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,1,32,power_law_1.2,1.1742784118652343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,2,16,balanced,0.1046720027923584
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,2,16,balanced,0.10605440139770508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,2,16,balanced,0.10933247566223145
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,2,16,balanced,0.021470720767974856
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,2,16,balanced,0.11985152244567872
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,2,16,balanced,0.12840191841125487
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,2,16,balanced,0.15221887588500976
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,2,16,balanced,0.16199552536010742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,2,16,balanced,0.2192255973815918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,2,16,balanced,0.25323135375976563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,1,32,power_law_1.2,0.903842544555664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,2,16,balanced,0.3459020614624023
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,2,16,balanced,0.023471360206604
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,2,16,balanced,0.4204646301269531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,2,16,balanced,0.02360192060470581
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,1,32,power_law_1.2,1.990443572998047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,2,16,balanced,0.02523391962051392
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,2,16,balanced,0.03641983985900879
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,2,16,balanced,0.6122534561157227
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,2,16,balanced,0.05995264053344727
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,2,16,balanced,0.05944064140319825
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,2,16,balanced,0.061711359024047854
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,2,16,balanced,0.06483456134796142
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,2,16,balanced,0.06529280185699463
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,2,16,balanced,0.0688704013824463
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,2,16,balanced,0.06209663867950439
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,2,16,balanced,0.06532864093780517
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,2,16,balanced,0.066909441947937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,2,16,balanced,0.07026048183441162
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,2,16,balanced,0.07554815769195557
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,2,16,balanced,0.08565247535705567
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,2,16,balanced,0.1001318359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,2,16,balanced,0.11312512397766113
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,2,16,balanced,0.10238080024719239
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,2,16,balanced,0.14325119972229003
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,2,16,balanced,0.17007360458374024
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,2,16,balanced,0.24178560256958007
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,2,16,balanced,0.0616703987121582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,2,16,balanced,0.7836914825439453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,2,16,balanced,0.29503999710083006
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,2,16,power_law_1.01,0.04729216098785401
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,2,16,balanced,0.42593921661376954
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,2,16,power_law_1.01,0.06962560176849365
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,2,16,power_law_1.01,0.06975359916687011
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,2,16,balanced,0.5372351837158204
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,2,16,power_law_1.01,0.07084799766540527
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,2,16,power_law_1.01,0.08462719917297364
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,2,16,power_law_1.01,0.0892403221130371
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,2,16,power_law_1.01,0.09361536026000976
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,2,16,power_law_1.01,0.09617664337158202
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,2,16,power_law_1.01,0.09820287704467774
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,2,16,power_law_1.01,0.10173184394836425
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,2,16,power_law_1.01,0.10469759941101073
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,2,16,power_law_1.01,0.10387583732604981
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,2,16,power_law_1.01,0.10214271545410156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,2,16,balanced,0.05315584182739258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,2,16,power_law_1.01,0.10429311752319337
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,2,16,power_law_1.01,0.10854399681091309
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,2,16,power_law_1.01,0.11610879898071289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,2,16,power_law_1.01,0.13581952095031738
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,2,16,power_law_1.01,0.1622617530822754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,2,16,power_law_1.01,0.19060607910156252
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,2,16,power_law_1.01,0.22337791442871097
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,2,16,power_law_1.01,0.292093448638916
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,2,16,power_law_1.01,0.06937600135803222
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,2,16,power_law_1.01,0.3693695831298828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,2,16,power_law_1.01,0.48863231658935546
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,2,16,power_law_1.01,0.050183677673339845
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,2,16,balanced,0.024080638885498044
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,2,16,power_law_1.01,0.05089280128479003
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,2,16,power_law_1.01,0.7071347045898437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,2,16,power_law_1.01,0.051814398765563964
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,2,16,power_law_1.01,0.05254655838012695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,2,16,power_law_1.01,0.05323775768280029
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,2,16,power_law_1.01,0.05688576221466064
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,2,16,power_law_1.01,0.05858431816101074
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,2,16,power_law_1.01,0.060143361091613765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,2,16,power_law_1.01,0.9745305633544922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,2,16,power_law_1.01,0.0614412784576416
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,2,16,power_law_1.01,0.06454144001007081
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,2,16,power_law_1.01,0.06558720111846925
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,2,16,power_law_1.01,0.06956160068511963
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,2,16,power_law_1.01,0.0625497579574585
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,2,16,power_law_1.01,0.06595327854156494
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,2,16,power_law_1.01,0.06782976150512696
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,2,16,power_law_1.01,0.07083136081695557
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,2,16,power_law_1.01,0.08524288177490234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,2,16,power_law_1.01,0.09364607810974121
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,2,16,power_law_1.01,0.03533056020736695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,2,16,power_law_1.01,0.11830400466918944
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,2,16,power_law_1.01,0.14553215980529785
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,2,16,power_law_1.01,0.1888947105407715
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,2,16,power_law_1.01,0.22780927658081054
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,2,16,power_law_1.01,0.3019046401977539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,2,16,power_law_1.2,0.04752511978149414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,2,16,power_law_1.01,0.4351551818847656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,2,16,power_law_1.2,0.06253056049346924
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,2,16,power_law_1.2,0.06941696166992187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,2,16,power_law_1.01,0.666519012451172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,2,16,power_law_1.01,1.282165832519531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,2,16,power_law_1.01,0.7191795349121094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,2,16,power_law_1.2,0.08513919830322265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,2,16,power_law_1.2,0.08922240257263184
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,2,16,power_law_1.2,0.09118847846984864
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,2,16,power_law_1.2,0.09434240341186524
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,2,16,power_law_1.2,0.09788031578063965
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,2,16,power_law_1.2,0.10399871826171876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,2,16,power_law_1.2,0.10810367584228515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,2,16,power_law_1.2,0.10963071823120116
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,2,16,power_law_1.2,0.10180992126464843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,2,16,power_law_1.2,0.10692352294921874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,2,16,power_law_1.2,0.109617919921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,2,16,power_law_1.2,0.1221337604522705
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,2,16,power_law_1.2,0.14315648078918458
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,2,16,power_law_1.2,0.0659609603881836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,2,16,power_law_1.2,0.17036415100097657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,2,16,power_law_1.2,0.203505916595459
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,2,16,power_law_1.2,0.27431808471679686
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,2,16,power_law_1.2,0.07061759948730469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,2,16,power_law_1.2,0.3603123092651367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,2,16,power_law_1.2,0.43043582916259765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,2,16,power_law_1.2,0.03525504112243653
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,2,16,power_law_1.2,0.05048448085784912
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,2,16,power_law_1.2,0.048299517631530765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,2,16,power_law_1.2,0.049328641891479494
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,2,16,power_law_1.2,0.811162872314453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,2,16,power_law_1.2,0.05227263927459717
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,2,16,power_law_1.2,0.04898943901062012
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,2,16,power_law_1.2,0.05720704078674317
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,2,16,power_law_1.2,0.059726080894470214
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,2,16,power_law_1.2,0.057372159957885735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,2,16,power_law_1.2,0.061674242019653325
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,2,16,power_law_1.2,1.2246797180175781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,2,16,power_law_1.2,0.06392704010009766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,2,16,power_law_1.2,0.06569727897644043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,2,16,power_law_1.2,0.06774911880493165
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,2,16,power_law_1.2,0.06368896007537841
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,2,16,power_law_1.2,0.06502143859863281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,2,16,power_law_1.2,0.07491072177886962
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,2,16,power_law_1.2,0.08657279968261719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,2,16,power_law_1.2,0.09814784049987793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,2,16,power_law_1.2,0.6145548629760742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,2,16,power_law_1.2,0.13801600456237792
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,2,16,power_law_1.2,0.22878719329833985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,2,16,power_law_1.2,0.2698457527160644
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,2,16,power_law_1.2,0.3787583923339844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,4,8,balanced,0.021200640201568605
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,4,8,balanced,0.02206592082977295
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,4,8,balanced,0.028368639945983886
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,4,8,balanced,0.039614720344543455
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,4,8,balanced,0.05068799972534179
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,2,16,power_law_1.2,0.791488037109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,2,16,power_law_1.2,0.06833024024963379
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,4,8,balanced,0.10199551582336426
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,4,8,balanced,0.10520832061767578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,2,16,power_law_1.2,0.16714752197265625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,4,8,balanced,0.10710271835327148
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,4,8,balanced,0.1096243190765381
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,2,16,power_law_1.2,0.9340595245361328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,4,8,balanced,0.10221952438354491
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,4,8,balanced,0.10285183906555176
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,4,8,balanced,0.10445055961608887
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,4,8,balanced,0.10746239662170409
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,4,8,balanced,0.1110041618347168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,4,8,balanced,0.11407744407653808
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,4,8,balanced,0.11995903968811035
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,4,8,balanced,0.13560832023620606
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,4,8,balanced,0.14873472213745118
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,4,8,balanced,0.18053632736206054
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,2,16,power_law_1.2,0.49185150146484374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,4,8,balanced,0.27548543930053715
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,4,8,balanced,0.3320985412597656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,4,8,balanced,0.07638527870178222
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,4,8,balanced,0.02339711904525757
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,2,16,power_law_1.2,1.610206756591797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,4,8,balanced,0.023733758926391603
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,4,8,balanced,0.5839244842529296
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,4,8,balanced,0.024396800994873048
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,4,8,balanced,0.027802879810333254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,4,8,balanced,0.03958911895751953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,4,8,balanced,0.06632063865661622
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,4,8,balanced,0.0579148817062378
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,4,8,balanced,0.8547392272949219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,4,8,balanced,0.0684223985671997
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,4,8,balanced,0.0706611204147339
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,4,8,balanced,0.07292799949645996
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,4,8,balanced,0.06511744022369384
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,4,8,balanced,0.06618112087249756
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,4,8,balanced,0.06857984066009522
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,4,8,balanced,1.1045388793945312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,4,8,balanced,0.07264383792877196
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,4,8,balanced,0.0746611213684082
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,4,8,balanced,0.0787059211730957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,4,8,balanced,0.08464768409729004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,4,8,balanced,0.09781760215759278
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,4,8,balanced,0.11066240310668944
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,4,8,balanced,0.46985214233398437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,4,8,balanced,0.16043392181396485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,4,8,balanced,0.20956544876098632
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,4,8,balanced,0.2579520034790039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,4,8,balanced,0.37433216094970706
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,4,8,power_law_1.01,0.03825279951095581
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,4,8,balanced,0.46940158843994145
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,4,8,power_law_1.01,0.04805376052856446
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,4,8,balanced,0.6887206268310547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,4,8,power_law_1.01,0.056861438751220704
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,4,8,power_law_1.01,0.06045567989349365
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,4,8,power_law_1.01,0.06676095962524414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,4,8,balanced,0.8876646423339845
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,4,8,power_law_1.01,0.093570556640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,4,8,power_law_1.01,0.09698047637939453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,4,8,balanced,0.1370150375366211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,4,8,power_law_1.01,0.09736191749572753
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,4,8,power_law_1.01,0.10139391899108888
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,4,8,power_law_1.01,0.10578304290771486
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,4,8,power_law_1.01,0.10496383666992189
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,4,8,power_law_1.01,0.09645952224731445
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,4,8,power_law_1.01,0.10130559921264648
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,4,8,power_law_1.01,0.10402688026428222
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,4,8,power_law_1.01,0.10815744400024414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,4,8,power_law_1.01,0.11330176353454591
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,4,8,power_law_1.01,0.1254220771789551
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,4,8,power_law_1.01,0.14737024307250976
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,4,8,power_law_1.01,0.17788415908813476
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,4,8,balanced,0.19877248764038086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,4,8,power_law_1.01,0.22004863739013675
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,4,8,power_law_1.01,0.301474552154541
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,4,8,power_law_1.01,0.3596121597290039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,4,8,power_law_1.01,0.06870912075042725
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,4,8,power_law_1.01,0.02520576000213623
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,4,8,power_law_1.01,0.4856217575073242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,4,8,power_law_1.01,0.03833215951919556
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,4,8,power_law_1.01,0.043137278556823734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,4,8,power_law_1.01,0.627191047668457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,4,8,power_law_1.01,0.046649599075317384
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,4,8,power_law_1.01,0.05141759872436523
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,4,8,power_law_1.01,0.05156223773956299
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,4,8,power_law_1.01,0.06356224060058593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,4,8,power_law_1.01,0.06354432106018067
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,4,8,power_law_1.01,0.0633996820449829
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,4,8,power_law_1.01,0.9058099365234374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,4,8,power_law_1.01,0.06770944118499755
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,4,8,power_law_1.01,0.07034239768981934
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,4,8,power_law_1.01,0.06351232051849365
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,4,8,power_law_1.01,0.06578303813934326
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,4,8,power_law_1.01,1.1537100982666015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,4,8,power_law_1.01,0.07110655784606934
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,4,8,power_law_1.01,0.07698048114776611
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,4,8,power_law_1.01,0.08931327819824218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,4,8,power_law_1.01,0.10379263877868652
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,4,8,power_law_1.01,0.12874367713928223
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,4,8,power_law_1.01,0.16545408248901367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,4,8,power_law_1.01,0.19688959121704103
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,4,8,power_law_1.01,0.24710271835327147
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,4,8,power_law_1.01,0.325662727355957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,4,8,power_law_1.2,0.03794048070907593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,4,8,power_law_1.01,0.4379046249389648
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,4,8,power_law_1.2,0.04664576053619385
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,4,8,power_law_1.2,0.04761343955993652
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,4,8,power_law_1.2,0.05611648082733154
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,4,8,power_law_1.01,0.6269696044921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,4,8,power_law_1.01,0.06108799934387207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,4,8,power_law_1.2,0.06936319828033446
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,4,8,power_law_1.2,0.09351167678833008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,4,8,power_law_1.2,0.09623167991638185
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,4,8,power_law_1.01,0.7949068450927734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,4,8,power_law_1.2,0.10017663955688476
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,4,8,power_law_1.2,0.10741760253906249
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,4,8,power_law_1.2,0.101976318359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,4,8,power_law_1.2,0.09756287574768066
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,4,8,power_law_1.2,0.10485504150390626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,4,8,power_law_1.2,0.10608384132385254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,4,8,power_law_1.2,0.1142745590209961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,4,8,power_law_1.2,0.13460351943969726
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,4,8,power_law_1.2,0.15519743919372558
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,4,8,power_law_1.2,0.18532480239868163
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,4,8,power_law_1.01,0.0656499195098877
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,4,8,power_law_1.2,0.23308927536010743
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,4,8,power_law_1.2,0.31980672836303714
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,4,8,power_law_1.2,0.07046144008636475
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,4,8,power_law_1.2,0.38721790313720705
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,4,8,power_law_1.2,0.10101375579833985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,4,8,power_law_1.2,0.025555200576782226
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,4,8,power_law_1.2,0.5651827239990235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,4,8,power_law_1.2,0.09999103546142578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,4,8,power_law_1.2,0.03817728042602539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,4,8,power_law_1.2,0.6684851074218751
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,4,8,power_law_1.2,0.03794559955596924
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,4,8,power_law_1.2,0.04305408000946045
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,4,8,power_law_1.2,0.05080192089080811
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,4,8,power_law_1.2,0.06074880123138428
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,4,8,power_law_1.2,0.0636134386062622
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,4,8,power_law_1.2,1.0606118774414062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,4,8,power_law_1.2,0.06523776054382324
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,4,8,power_law_1.2,0.06559360027313232
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,4,8,power_law_1.2,0.06945792198181153
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,4,8,power_law_1.2,0.07085311889648438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,4,8,power_law_1.2,0.06198783874511719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,4,8,power_law_1.2,0.06339456081390381
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,4,8,power_law_1.2,1.2143936157226562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,4,8,power_law_1.2,0.06754687786102295
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,4,8,power_law_1.2,0.06988800048828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,4,8,power_law_1.2,0.07901567935943603
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,4,8,power_law_1.2,0.09636735916137695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,4,8,power_law_1.2,0.10699007987976075
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,4,8,power_law_1.2,0.12844415664672854
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,4,8,power_law_1.2,0.15852928161621094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,4,8,power_law_1.2,0.21983871459960938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,4,8,power_law_1.2,0.28711168289184574
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,8,4,balanced,0.020606720447540285
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,4,8,power_law_1.2,0.3774054336547851
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,8,4,balanced,0.02092416048049927
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,8,4,balanced,0.028501760959625245
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,8,4,balanced,0.03930111885070801
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,8,4,balanced,0.04977536201477051
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,8,4,balanced,0.07306879997253418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,8,4,balanced,0.07501440048217774
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,4,8,power_law_1.2,0.7196185302734375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,8,4,balanced,0.07585919857025146
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,8,4,balanced,0.09649151802062989
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,8,4,balanced,0.09694592475891113
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,8,4,balanced,0.10025088310241699
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,8,4,balanced,0.09856127738952637
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,8,4,balanced,0.10185215950012208
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,8,4,balanced,0.10431488037109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,8,4,balanced,0.1071897602081299
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,8,4,balanced,0.11024767875671386
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,8,4,balanced,0.11577728271484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,8,4,balanced,0.12943103790283203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,8,4,balanced,0.14007040023803713
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,8,4,balanced,0.1863577651977539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,4,8,power_law_1.2,0.04604415893554688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,8,4,balanced,0.2672204780578613
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,4,8,power_law_1.2,0.46755329132080076
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,8,4,balanced,0.3199423980712891
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,8,4,balanced,0.022498559951782227
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,8,4,balanced,0.022928640842437745
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,8,4,balanced,0.461767692565918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,8,4,balanced,0.02471679925918579
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,8,4,balanced,0.02908031940460205
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,4,8,power_law_1.2,0.8668339538574219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,8,4,balanced,0.06530943870544434
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,8,4,balanced,0.06410880088806152
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,8,4,balanced,0.07712639808654785
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,8,4,balanced,0.07737855911254883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,8,4,balanced,0.1640870475769043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,8,4,balanced,0.07870207786560059
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,8,4,balanced,0.08054528236389161
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,8,4,balanced,0.08229632377624511
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,8,4,balanced,0.08527615547180176
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,8,4,balanced,0.08701824188232422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,8,4,balanced,0.08800383567810058
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,8,4,balanced,0.09431039810180664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,8,4,balanced,0.10715776443481446
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,8,4,balanced,0.11601920127868652
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,8,4,balanced,0.042120962142944335
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,8,4,balanced,0.14179327964782715
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,8,4,balanced,0.063438081741333
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,8,4,balanced,0.2070361518859863
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,8,4,balanced,0.2523507118225098
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,8,4,balanced,0.37895809173583983
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,8,4,balanced,1.0933683013916016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,8,4,balanced,0.46809726715087885
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,8,4,power_law_1.01,0.02166912078857422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,8,4,power_law_1.01,0.03815040111541748
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,8,4,balanced,0.6833945465087891
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,8,4,power_law_1.01,0.048568320274353025
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,8,4,balanced,0.5802508926391601
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,8,4,power_law_1.01,0.05471871852874756
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,8,4,power_law_1.01,0.054688000679016115
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,8,4,power_law_1.01,0.0684441614151001
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,8,4,power_law_1.01,0.06999040126800538
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,8,4,power_law_1.01,0.0928435230255127
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,8,4,power_law_1.01,0.09231743812561036
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,8,4,power_law_1.01,0.09559295654296876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,8,4,power_law_1.01,0.09576576232910157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,8,4,power_law_1.01,0.10074879646301269
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,8,4,power_law_1.01,0.10270336151123047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,8,4,power_law_1.01,0.10627072334289552
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,8,4,power_law_1.01,0.11180543899536133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,8,4,power_law_1.01,0.04298495769500733
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,8,4,power_law_1.01,0.12913663864135744
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,8,4,power_law_1.01,0.1416985607147217
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,8,4,power_law_1.01,0.17250944137573243
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,8,4,balanced,0.8996339416503906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,8,4,balanced,0.8446630096435547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,8,4,power_law_1.01,0.2819340705871582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,8,4,power_law_1.01,0.0939635181427002
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,8,4,power_law_1.01,0.3563955307006836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,8,4,power_law_1.01,0.4926591873168945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,8,4,power_law_1.01,0.02398591995239258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,8,4,power_law_1.01,0.623870735168457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,8,4,power_law_1.01,0.027263998985290527
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,8,4,power_law_1.01,0.033323519229888916
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,8,4,power_law_1.01,0.8940480041503907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,8,4,power_law_1.01,0.040144639015197756
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,8,4,power_law_1.01,0.21156864166259765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,8,4,power_law_1.01,0.04693247795104981
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,8,4,power_law_1.01,0.046616320610046384
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,8,4,power_law_1.01,0.05933184146881103
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,8,4,power_law_1.01,1.172375030517578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,8,4,power_law_1.01,0.07220992088317871
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,8,4,power_law_1.01,0.06895872116088866
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,8,4,power_law_1.01,0.07383935928344727
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,8,4,power_law_1.01,0.07363584041595458
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,8,4,power_law_1.01,0.07492991924285888
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,8,4,power_law_1.01,0.0784934377670288
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,8,4,power_law_1.01,0.08223999977111816
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,8,4,power_law_1.01,0.08913023948669434
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,8,4,power_law_1.01,0.10475647926330567
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,8,4,power_law_1.01,0.12672512054443358
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,8,4,power_law_1.01,0.15492480278015136
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,8,4,balanced,0.16235008239746093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,8,4,power_law_1.01,0.223187198638916
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,8,4,power_law_1.01,0.2720998382568359
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,8,4,power_law_1.01,0.05907584190368652
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,8,4,power_law_1.01,0.39698814392089843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,8,4,power_law_1.2,0.02185215950012207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,8,4,power_law_1.2,0.03700223922729492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,8,4,power_law_1.01,0.08398336410522461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,8,4,power_law_1.2,0.04033664226531982
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,8,4,power_law_1.2,0.046417918205261234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,8,4,power_law_1.2,0.05335807800292969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,8,4,power_law_1.2,0.05250432014465332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,8,4,power_law_1.2,0.07162240028381348
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,8,4,power_law_1.01,0.18010112762451172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,8,4,power_law_1.2,0.09097984313964844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,8,4,power_law_1.01,0.9506585693359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,8,4,power_law_1.2,0.09283840179443359
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,8,4,power_law_1.2,0.09400447845458984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,8,4,power_law_1.2,0.09587712287902832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,8,4,power_law_1.2,0.09985152244567871
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,8,4,power_law_1.2,0.10249855995178223
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,8,4,power_law_1.2,0.10607616424560547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,8,4,power_law_1.2,0.11401087760925292
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,8,4,power_law_1.2,0.12945280075073243
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,8,4,power_law_1.01,0.4704051208496094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,8,4,power_law_1.2,0.14297216415405273
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,8,4,power_law_1.2,0.17725311279296876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,8,4,power_law_1.2,0.21876863479614256
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,8,4,power_law_1.2,0.3000268745422363
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,8,4,power_law_1.2,0.07273215770721435
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,8,4,power_law_1.2,0.37777793884277344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,8,4,power_law_1.2,0.024156160354614258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,8,4,power_law_1.2,0.5103564834594726
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,8,4,power_law_1.2,0.09765119552612304
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,8,4,power_law_1.2,0.028579840660095213
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,8,4,power_law_1.2,0.6776525115966796
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,8,4,power_law_1.2,0.0326528000831604
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,8,4,power_law_1.2,0.0392576003074646
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,8,4,power_law_1.2,0.04586624145507813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,8,4,power_law_1.2,0.04537856101989746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,8,4,power_law_1.2,0.06270336151123047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,8,4,power_law_1.2,0.059202561378479
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,8,4,power_law_1.2,0.9766617584228516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,8,4,power_law_1.2,0.0700044822692871
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,8,4,power_law_1.2,0.07395328044891357
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,8,4,power_law_1.01,0.7265190124511719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,8,4,power_law_1.2,0.0763366413116455
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,8,4,power_law_1.2,0.0754585599899292
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,8,4,power_law_1.2,0.07933440208435058
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,8,4,power_law_1.2,1.223279342651367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,8,4,power_law_1.2,0.08308735847473145
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,8,4,power_law_1.2,0.0926848030090332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,8,4,power_law_1.2,0.1184115219116211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,8,4,power_law_1.2,0.1327526378631592
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,8,4,power_law_1.2,0.15871999740600587
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,8,4,power_law_1.2,0.18475135803222656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,8,4,power_law_1.2,0.24034944534301758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,16,2,balanced,0.020514559745788575
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,8,4,power_law_1.2,0.413623046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,16,2,balanced,0.021996800899505613
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,8,4,power_law_1.2,0.06885632038116454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,8,4,power_law_1.2,0.5364492797851563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,16,2,balanced,0.03088639974594116
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,16,2,balanced,0.057310719490051266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,16,2,balanced,0.08621439933776856
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,16,2,balanced,0.08577664375305176
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,8,4,power_law_1.2,0.7721663665771484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,8,4,power_law_1.2,0.08532480239868165
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,16,2,balanced,0.08670975685119628
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,16,2,balanced,0.0879526424407959
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,16,2,balanced,0.08811391830444336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,16,2,balanced,0.10379391670227052
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,8,4,power_law_1.2,0.9951679992675782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,16,2,balanced,0.1050175952911377
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,16,2,balanced,0.10671744346618653
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,16,2,balanced,0.11037952423095702
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,16,2,balanced,0.11263615608215331
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,8,4,power_law_1.2,0.2990937614440918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,16,2,balanced,0.12928256034851074
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,16,2,balanced,0.1392319965362549
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,16,2,balanced,0.1631590461730957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,16,2,balanced,0.18327167510986328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,16,2,balanced,0.25845632553100584
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,16,2,balanced,0.04100607872009278
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,16,2,balanced,0.3156492805480957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,16,2,balanced,0.022442240715026856
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,16,2,balanced,0.4604044723510742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,16,2,balanced,0.089684476852417
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,16,2,balanced,0.023948800563812257
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,16,2,balanced,0.5916723251342774
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,16,2,balanced,0.02659584045410156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,16,2,balanced,0.03344255924224854
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,16,2,balanced,0.051041278839111325
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,16,2,balanced,0.08019200325012207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,16,2,balanced,0.08316415786743164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,16,2,balanced,0.8663091278076172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,16,2,balanced,0.08124799728393554
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,16,2,balanced,0.08178048133850098
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,16,2,balanced,0.08203776359558106
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,16,2,balanced,0.0843942356109619
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,16,2,balanced,0.10514816284179687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,16,2,balanced,0.1045081615447998
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,16,2,balanced,0.10768896102905275
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,16,2,balanced,0.10881152153015136
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,16,2,balanced,0.11227904319763185
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,16,2,balanced,0.11657216072082519
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,16,2,balanced,0.1270143985748291
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,16,2,balanced,0.13359359741210936
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,16,2,balanced,0.15631360054016114
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,16,2,balanced,0.1747724723815918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,16,2,balanced,0.2160665512084961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,16,2,balanced,0.25677568435668946
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,16,2,balanced,0.3950374221801758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,16,2,balanced,0.11738495826721192
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,16,2,balanced,0.4784281539916992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,16,2,power_law_1.01,0.02372607946395874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,16,2,power_law_1.01,0.03758591890335083
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,16,2,balanced,0.703994903564453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,16,2,power_law_1.01,0.04449151992797852
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,16,2,balanced,1.1377049255371094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,16,2,power_law_1.01,0.054869761466979985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,16,2,power_law_1.01,0.06184063911437988
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,16,2,power_law_1.01,0.07239935874938966
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,16,2,power_law_1.01,0.0763327980041504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,16,2,power_law_1.01,0.07339263916015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,16,2,power_law_1.01,0.07869311809539795
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,16,2,power_law_1.01,0.09595392227172853
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,16,2,power_law_1.01,0.09684351921081544
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,16,2,power_law_1.01,0.10150400161743164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,16,2,power_law_1.01,0.10360447883605957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,16,2,power_law_1.01,0.10767871856689452
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,16,2,power_law_1.01,0.11363967895507812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,16,2,power_law_1.01,0.1283827209472656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,16,2,power_law_1.01,0.14031488418579102
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,16,2,power_law_1.01,0.16986879348754882
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,16,2,power_law_1.01,0.021413118839263917
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,16,2,power_law_1.01,0.19910144805908203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,16,2,power_law_1.01,0.2755699157714844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,16,2,power_law_1.01,0.3493824005126953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,16,2,power_law_1.01,0.023511040210723876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,16,2,power_law_1.01,0.48605697631835937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,16,2,power_law_1.01,0.08264063835144043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,16,2,power_law_1.01,0.6152844619750977
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,16,2,power_law_1.01,0.03228672027587891
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,16,2,power_law_1.01,0.0386406397819519
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,16,2,power_law_1.01,0.8705152130126953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,16,2,power_law_1.01,0.048014078140258785
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,16,2,power_law_1.01,0.05949952125549316
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,16,2,power_law_1.01,0.06509183883666993
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,16,2,power_law_1.01,0.0673689603805542
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,16,2,power_law_1.01,0.07138944149017334
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,16,2,power_law_1.01,0.07040128231048584
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,16,2,power_law_1.01,1.161443862915039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,16,2,power_law_1.01,0.09230208396911621
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,16,2,power_law_1.01,0.09306495666503907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,16,2,power_law_1.01,0.09749247550964354
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,16,2,power_law_1.01,0.09953536033630371
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,16,2,power_law_1.01,0.10324480056762694
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,16,2,power_law_1.01,0.10802047729492188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,16,2,power_law_1.01,0.12156543731689454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,16,2,balanced,0.9306585693359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,16,2,power_law_1.01,0.14058496475219726
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,16,2,power_law_1.01,0.026321918964385987
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,16,2,power_law_1.01,0.17657983779907227
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,16,2,power_law_1.01,0.1993612861633301
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,16,2,power_law_1.01,0.25374847412109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,16,2,power_law_1.01,0.3075187110900879
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,16,2,power_law_1.2,0.02126336097717285
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,16,2,power_law_1.2,0.030127360820770267
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,16,2,power_law_1.01,0.4258534240722656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,16,2,power_law_1.2,0.036720640659332275
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,16,2,power_law_1.01,0.07314303874969483
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,16,2,power_law_1.01,0.5384870529174804
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,16,2,power_law_1.2,0.05279488086700439
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,16,2,power_law_1.01,0.773207015991211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,16,2,power_law_1.2,0.06972928047180176
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,16,2,power_law_1.2,0.07237760066986085
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,16,2,power_law_1.01,0.9909490966796876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,16,2,power_law_1.2,0.07699327945709229
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,16,2,power_law_1.2,0.09631744384765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,16,2,power_law_1.2,0.10113535881042482
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,16,2,power_law_1.2,0.10430208206176758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,16,2,power_law_1.2,0.10896256446838379
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,16,2,power_law_1.2,0.11568256378173829
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,16,2,power_law_1.2,0.043610877990722656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,16,2,power_law_1.2,0.13114751815795897
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,16,2,power_law_1.2,0.06099199771881103
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,16,2,power_law_1.2,0.14438783645629882
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,16,2,power_law_1.2,0.17124223709106445
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,16,2,power_law_1.2,0.0756876802444458
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,16,2,power_law_1.2,0.08292863845825196
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,16,2,power_law_1.2,0.20655359268188475
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,16,2,power_law_1.2,0.2806502342224121
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,16,2,power_law_1.2,0.3605388641357422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,16,2,power_law_1.2,0.09796607971191407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,16,2,power_law_1.2,0.4944345474243164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,16,2,power_law_1.2,0.02341248035430908
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,16,2,power_law_1.2,0.026426880359649657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,16,2,power_law_1.2,0.03127295970916748
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,16,2,power_law_1.2,0.6278144073486328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,16,2,power_law_1.2,0.044846081733703615
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,16,2,power_law_1.2,0.051357440948486335
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,16,2,power_law_1.2,0.06431871891021729
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,16,2,power_law_1.2,0.8910412597656251
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,16,2,power_law_1.2,0.06824575901031495
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,16,2,power_law_1.2,0.06917888164520264
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,16,2,power_law_1.2,0.0702566385269165
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,16,2,power_law_1.2,0.07300608158111573
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,16,2,power_law_1.2,0.09286911964416504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,16,2,power_law_1.2,1.1945804595947265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,16,2,power_law_1.2,0.09464832305908202
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,16,2,power_law_1.2,0.1043507194519043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,16,2,power_law_1.2,0.11249024391174316
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,16,2,power_law_1.2,0.12972031593322755
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,16,2,power_law_1.2,0.1819980812072754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,16,2,power_law_1.2,0.150764799118042
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,16,2,power_law_1.2,0.20651008605957033
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,16,2,power_law_1.2,0.03802367925643921
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,16,2,power_law_1.2,0.26231807708740235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,16,2,power_law_1.2,0.32247550964355465
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,32,1,balanced,0.028305280208587646
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,16,2,power_law_1.2,0.4386214447021485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,32,1,balanced,0.033594241142272954
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,16,2,power_law_1.2,0.5560319900512696
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,16,2,power_law_1.2,0.09883904457092285
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,16,2,power_law_1.2,0.10160639762878418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,32,1,balanced,0.11481535911560059
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,32,1,balanced,0.1156230354309082
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,16,2,power_law_1.2,0.7974822235107422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,32,1,balanced,0.11682880401611329
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,32,1,balanced,0.11812159538269043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,16,2,power_law_1.2,1.0073638153076172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,32,1,balanced,0.12307647705078124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,32,1,balanced,0.12498559951782226
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,32,1,balanced,0.1304537582397461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,32,1,balanced,0.13551551818847657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,32,1,balanced,0.13630656242370603
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,32,1,balanced,0.0416153621673584
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,32,1,balanced,0.1411961555480957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,32,1,balanced,0.054675841331481935
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,32,1,balanced,0.1500096035003662
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,32,1,balanced,0.1565824031829834
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,32,1,balanced,0.17951744079589843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,32,1,balanced,0.20337600708007814
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,32,1,balanced,0.1162553596496582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,32,1,balanced,0.29826303482055666
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,32,1,balanced,0.3391379165649414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,32,1,balanced,0.12028672218322753
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,32,1,balanced,0.03085119962692261
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,32,1,balanced,0.0343180799484253
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,32,1,balanced,0.490819206237793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,32,1,balanced,0.039312639236450196
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,32,1,balanced,0.051312642097473146
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,32,1,balanced,0.07815872192382813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,32,1,balanced,0.12938303947448732
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,32,1,balanced,0.6504793548583985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,32,1,balanced,0.12868608474731444
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,32,1,balanced,0.0754265594482422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,32,1,balanced,0.13140607833862306
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,32,1,balanced,0.13208064079284668
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,32,1,balanced,0.1336473560333252
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,32,1,balanced,0.13515392303466797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,32,1,balanced,0.13711615562438967
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,32,1,balanced,0.1394655990600586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,32,1,balanced,0.18039680480957032
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,32,1,balanced,0.18260736465454103
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,32,1,balanced,0.19374975204467773
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,32,1,balanced,1.2923301696777343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,32,1,balanced,0.22068159103393556
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,32,1,balanced,0.23782720565795898
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,32,1,balanced,0.27815040588378903
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,32,1,balanced,0.13096896171569825
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,32,1,balanced,0.9721638488769532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,32,1,balanced,0.515601921081543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,32,1,balanced,0.18437311172485352
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,32,1,balanced,0.5987206268310546
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,32,1,balanced,0.20076736450195312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,32,1,power_law_1.01,0.027599999904632567
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,32,1,power_law_1.01,0.031098239421844483
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,32,1,power_law_1.01,0.04521215915679931
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,32,1,balanced,0.8802054595947265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,32,1,balanced,0.31710975646972656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,32,1,power_law_1.01,0.06702400207519531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,32,1,power_law_1.01,0.08208319664001465
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,32,1,power_law_1.01,0.09758272171020507
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,32,1,power_law_1.01,0.10101311683654786
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,32,1,power_law_1.01,0.10465536117553711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,32,1,power_law_1.01,0.109169921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,32,1,balanced,1.1604518127441406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,32,1,power_law_1.01,0.11574591636657713
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,32,1,power_law_1.01,0.12009599685668945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,32,1,power_law_1.01,0.12926719665527345
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,32,1,power_law_1.01,0.12748224258422852
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,32,1,power_law_1.01,0.12961343765258787
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,32,1,power_law_1.01,0.13655424118041992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,32,1,power_law_1.01,0.15106176376342773
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,32,1,power_law_1.01,0.1639379119873047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,32,1,power_law_1.01,0.1958099174499512
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,32,1,power_law_1.01,0.22957376480102537
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,32,1,power_law_1.01,0.3168000030517578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,32,1,power_law_1.01,0.10234560012817381
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,32,1,power_law_1.01,0.39329406738281253
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,32,1,power_law_1.01,0.031004159450531005
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,32,1,power_law_1.01,0.0340230393409729
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,32,1,power_law_1.01,0.533735694885254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,32,1,power_law_1.01,0.0409830379486084
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,32,1,power_law_1.01,0.6939110565185547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,32,1,power_law_1.01,0.05468416213989258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,32,1,power_law_1.01,0.05377151966094971
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,32,1,power_law_1.01,0.066942720413208
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,32,1,power_law_1.01,0.0830720043182373
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,32,1,power_law_1.01,1.0033740997314453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,32,1,power_law_1.01,0.1037881565093994
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,32,1,power_law_1.01,0.10758399963378906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,32,1,power_law_1.01,0.10993087768554688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,32,1,power_law_1.01,0.11123519897460939
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,32,1,power_law_1.01,0.11550975799560548
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,32,1,power_law_1.01,0.11940095901489259
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,32,1,power_law_1.01,0.12405504226684569
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,32,1,power_law_1.01,0.12115776062011718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,32,1,power_law_1.01,0.1676121520996094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,32,1,power_law_1.01,0.17742143630981447
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,32,1,power_law_1.01,1.321329345703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,32,1,power_law_1.01,0.18907199859619142
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,32,1,power_law_1.01,0.20854656219482423
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,32,1,power_law_1.01,0.2498956871032715
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,32,1,power_law_1.01,0.2794521522521973
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,32,1,power_law_1.01,0.3443475341796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,32,1,power_law_1.01,0.4241632080078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,32,1,power_law_1.2,0.027534079551696778
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,32,1,power_law_1.01,0.5561568069458007
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,32,1,power_law_1.2,0.045293440818786626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,32,1,power_law_1.01,0.16662399291992186
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,32,1,power_law_1.01,0.7090879821777344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,32,1,power_law_1.2,0.05332223892211914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,32,1,power_law_1.2,0.06419648170471191
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,32,1,power_law_1.2,0.08032896041870116
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,32,1,power_law_1.2,0.10023103713989259
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,32,1,power_law_1.01,0.9907411193847656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,32,1,power_law_1.2,0.10191743850708009
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,32,1,power_law_1.2,0.10454655647277833
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,32,1,power_law_1.2,0.11586496353149414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,32,1,power_law_1.2,0.1209670352935791
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,32,1,power_law_1.01,1.2681478118896483
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,32,1,power_law_1.2,0.031064960956573483
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,32,1,power_law_1.2,0.12897600173950194
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,32,1,power_law_1.2,0.1326476764678955
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,32,1,power_law_1.2,0.1387788772583008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,32,1,power_law_1.2,0.154968957901001
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,32,1,power_law_1.2,0.19982336044311524
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,32,1,power_law_1.2,0.09659903526306153
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,32,1,power_law_1.2,0.23410303115844727
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,32,1,power_law_1.2,0.3269977569580078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,32,1,power_law_1.2,0.10976063728332519
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,32,1,power_law_1.2,0.39999679565429686
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,32,1,power_law_1.2,0.1317094421386719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,32,1,power_law_1.2,0.030935039520263673
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,32,1,power_law_1.2,0.5420652770996094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,32,1,power_law_1.2,0.03427135944366455
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,32,1,power_law_1.2,0.04053760051727295
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,32,1,power_law_1.2,0.16734655380249025
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,32,1,power_law_1.2,0.7034585571289063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,32,1,power_law_1.2,0.05283008098602295
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,32,1,power_law_1.2,0.06415743827819824
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,32,1,power_law_1.2,0.08258048057556153
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,32,1,power_law_1.2,0.1019820785522461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,32,1,power_law_1.2,0.10873536109924316
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,32,1,power_law_1.2,1.018202896118164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,32,1,power_law_1.2,0.11163392066955566
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,32,1,power_law_1.2,0.11451647758483888
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,32,1,power_law_1.2,0.11821760177612303
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,32,1,power_law_1.2,0.12148223876953126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,32,1,power_law_1.2,0.1262502384185791
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,32,1,power_law_1.2,0.1681964874267578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,32,1,power_law_1.2,1.3414924621582032
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,32,1,power_law_1.2,0.17357824325561524
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,32,1,power_law_1.2,0.17871103286743165
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,32,1,power_law_1.2,0.20027328491210938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,32,1,power_law_1.2,0.22690879821777346
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,32,1,power_law_1.2,0.25938943862915037
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,32,1,power_law_1.2,0.10644927978515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,32,1,power_law_1.2,0.2885785675048828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,32,1,power_law_1.2,0.3582566452026367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,1,32,balanced,0.023301119804382327
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,1,32,balanced,0.023854079246520995
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,1,32,balanced,0.02313983917236328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,1,32,balanced,0.046680321693420415
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,1,32,balanced,0.07412864208221435
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,32,1,power_law_1.2,0.5663270568847656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,1,32,balanced,0.07919616222381591
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,1,32,balanced,0.08107392311096191
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,32,1,power_law_1.2,0.718089599609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,1,32,balanced,0.0813913631439209
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,1,32,balanced,0.08112128257751465
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,1,32,balanced,0.08202752113342285
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,1,32,balanced,0.08197759628295899
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,1,32,balanced,0.08315903663635253
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,1,32,balanced,0.08653823852539062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,1,32,balanced,0.08623104095458985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,1,32,balanced,0.0895628833770752
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,1,32,balanced,0.09118592262268066
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,1,32,balanced,0.09435008049011231
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,1,32,balanced,0.10093055725097658
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,1,32,balanced,0.09772671699523926
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,1,32,balanced,0.1014246368408203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,1,32,balanced,0.11248640060424804
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,1,32,balanced,0.14061568260192872
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,1,32,balanced,0.1499673557281494
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,1,32,balanced,0.2168435287475586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,32,1,power_law_1.2,1.2870918273925782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,1,32,balanced,0.25063552856445315
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,1,32,power_law_1.01,0.05144447803497314
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,1,32,power_law_1.01,0.05075583934783936
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,1,32,power_law_1.01,0.051466240882873535
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,1,32,balanced,0.4507737731933593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,1,32,power_law_1.01,0.050530557632446294
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,1,32,power_law_1.01,0.07727744102478026
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,1,32,power_law_1.01,0.07970431804656983
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,32,1,power_law_1.2,0.9974463653564453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,1,32,power_law_1.01,0.08227071762084961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,1,32,power_law_1.01,0.08339839935302734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,1,32,power_law_1.01,0.08484607696533203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,1,32,power_law_1.01,0.09107199668884278
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,32,1,power_law_1.2,0.43412929534912104
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,1,32,power_law_1.01,0.09566720008850098
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,1,32,power_law_1.01,0.10002176284790039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,1,32,power_law_1.01,0.10024703979492186
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,1,32,power_law_1.01,0.10150272369384765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,1,32,power_law_1.01,0.10813952445983888
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,1,32,power_law_1.01,0.13501440048217775
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,1,32,balanced,0.35022079467773437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,1,32,power_law_1.01,0.17124223709106445
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,1,32,power_law_1.01,0.1993280029296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,1,32,power_law_1.01,0.2745702362060547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,1,32,power_law_1.01,0.08135552406311035
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,1,32,power_law_1.01,0.4030579376220703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,1,32,power_law_1.01,0.08828543663024903
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,1,32,power_law_1.01,0.5375603103637696
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,1,32,power_law_1.2,0.050832638740539546
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,1,32,power_law_1.2,0.050888957977294924
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,1,32,power_law_1.01,0.7369522857666015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,1,32,power_law_1.2,0.05138944149017334
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,1,32,power_law_1.2,0.05016448020935058
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,1,32,power_law_1.2,0.07557119846343993
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,1,32,power_law_1.2,0.08029696464538574
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,1,32,power_law_1.2,0.0819315242767334
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,1,32,power_law_1.2,0.0830720043182373
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,1,32,power_law_1.2,0.08608639717102051
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,1,32,power_law_1.01,1.0433650970458985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,1,32,power_law_1.2,0.08897536277770997
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,1,32,power_law_1.2,0.08963839530944824
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,1,32,power_law_1.2,0.09327487945556641
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,1,32,power_law_1.2,0.0970355224609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,1,32,power_law_1.2,0.10076160430908203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,1,32,power_law_1.2,0.10197504043579102
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,1,32,power_law_1.2,0.102225923538208
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,1,32,power_law_1.2,0.13361023902893066
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,1,32,power_law_1.2,0.15540096282958984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,1,32,power_law_1.01,1.4512818908691405
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,1,32,power_law_1.2,0.21710079193115234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,1,32,power_law_1.01,0.08737279891967772
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,1,32,power_law_1.2,0.27225088119506835
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,1,32,power_law_1.2,0.332250862121582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,2,16,balanced,0.020573439598083495
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,2,16,balanced,0.020742399692535402
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,1,32,power_law_1.2,0.5051788711547852
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,1,32,power_law_1.2,0.08048255920410156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,2,16,balanced,0.03813760042190552
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,2,16,balanced,0.043265280723571775
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,2,16,balanced,0.07059711933135986
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,2,16,balanced,0.07186816215515138
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,2,16,balanced,0.07193471908569335
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,2,16,balanced,0.07283967971801758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,2,16,balanced,0.07599999904632568
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,1,32,power_law_1.2,1.1845606231689454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,2,16,balanced,0.07732223987579345
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,2,16,balanced,0.07950719833374023
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,2,16,balanced,0.08117119789123535
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,2,16,balanced,0.08273792266845703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,2,16,balanced,0.08864895820617676
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,2,16,balanced,0.09076607704162598
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,2,16,balanced,0.08156415939331055
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,2,16,balanced,0.08366463661193849
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,2,16,balanced,0.08864383697509766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,2,16,balanced,0.10845696449279785
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,2,16,balanced,0.11586688041687013
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,1,32,power_law_1.2,0.7241766357421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,1,32,power_law_1.2,1.9396659851074218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,2,16,balanced,0.14352255821228027
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,2,16,balanced,0.16587263107299804
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,2,16,balanced,0.24434303283691405
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,2,16,balanced,0.28789888381958006
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,2,16,power_law_1.01,0.04218751907348633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,2,16,power_law_1.01,0.04790143966674805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,2,16,balanced,0.41071998596191406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,2,16,power_law_1.01,0.045701122283935545
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,2,16,balanced,0.5290508651733399
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,2,16,power_law_1.01,0.04700032234191894
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,2,16,power_law_1.01,0.06703360080718994
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,2,16,power_law_1.01,0.06872320175170898
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,2,16,power_law_1.01,0.06965760231018067
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,2,16,power_law_1.01,0.07505792140960693
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,2,16,power_law_1.01,0.07260672092437745
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,2,16,power_law_1.01,0.0759603214263916
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,2,16,power_law_1.01,0.07862271785736083
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,2,16,power_law_1.01,0.08121600151062011
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,2,16,power_law_1.01,0.08274175643920897
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,2,16,power_law_1.01,0.0872383975982666
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,2,16,power_law_1.01,0.08395008087158204
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,2,16,power_law_1.01,0.08727552413940429
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,2,16,power_law_1.01,0.08440192222595215
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,2,16,power_law_1.01,0.09504256248474122
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,2,16,power_law_1.01,0.10965375900268555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,2,16,power_law_1.01,0.1444057559967041
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,1,32,power_law_1.2,1.5567578125000001
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,2,16,power_law_1.01,0.17158784866333007
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,2,16,power_law_1.01,0.24310144424438476
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,2,16,power_law_1.01,0.3070835113525391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,2,16,power_law_1.2,0.04227583885192871
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,2,16,power_law_1.2,0.0471065616607666
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,2,16,power_law_1.01,0.4252979278564453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,2,16,balanced,0.021181440353393553
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,2,16,power_law_1.2,0.046983680725097655
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,2,16,power_law_1.2,0.046926078796386717
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,2,16,power_law_1.01,0.5523737716674805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,2,16,power_law_1.2,0.06795648097991944
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,2,16,power_law_1.2,0.06851200103759766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,2,16,power_law_1.2,0.06922880172729493
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,2,16,power_law_1.2,0.07438464164733886
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,2,16,power_law_1.01,0.7094566345214843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,2,16,power_law_1.2,0.07587071895599365
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,2,16,power_law_1.2,0.07754752159118652
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,2,16,power_law_1.2,0.0798579216003418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,2,16,power_law_1.2,0.08332544326782226
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,2,16,power_law_1.2,0.09111040115356446
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,2,16,power_law_1.2,0.08576383590698242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,2,16,power_law_1.2,0.08474240303039551
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,2,16,power_law_1.2,0.10841343879699708
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,2,16,power_law_1.2,0.12260607719421386
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,2,16,power_law_1.2,0.16151424407958986
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,2,16,power_law_1.2,0.1927039909362793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,2,16,power_law_1.2,0.28418560028076173
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,2,16,power_law_1.2,0.33339649200439453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,4,8,balanced,0.01931007981300354
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,2,16,power_law_1.2,0.491278076171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,2,16,power_law_1.2,0.07268735885620117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,4,8,balanced,0.020084478855133057
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,4,8,balanced,0.022430720329284667
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,4,8,balanced,0.035777280330657954
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,4,8,balanced,0.045401601791381835
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,4,8,balanced,0.06939648151397705
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,4,8,balanced,0.07341055870056153
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,4,8,balanced,0.07452799797058104
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,4,8,balanced,0.07681407928466796
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,2,16,power_law_1.2,0.9766515350341797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,4,8,balanced,0.07753087997436522
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,4,8,balanced,0.07920383930206298
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,4,8,balanced,0.08708224296569825
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,4,8,balanced,0.07997056007385253
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,4,8,balanced,0.08907008171081543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,4,8,balanced,0.08410240173339843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,4,8,balanced,0.08077568054199219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,4,8,balanced,0.08535552024841309
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,4,8,balanced,0.09291007995605469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,4,8,balanced,0.10347135543823242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,4,8,balanced,0.12302335739135742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,4,8,balanced,0.1408255958557129
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,4,8,balanced,0.18690176010131837
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,4,8,balanced,0.2204825592041016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,2,16,power_law_1.01,0.9729561614990235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,2,16,power_law_1.2,0.08391424179077148
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,4,8,balanced,0.389884147644043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,4,8,power_law_1.01,0.030280959606170655
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,4,8,balanced,0.5641638565063476
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,4,8,power_law_1.01,0.040668158531188964
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,4,8,power_law_1.01,0.039755520820617674
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,4,8,power_law_1.01,0.04547967910766602
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,4,8,power_law_1.01,0.04546048164367676
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,2,16,power_law_1.2,0.7617574310302734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,4,8,balanced,0.7376102447509766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,4,8,power_law_1.01,0.06578303813934326
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,4,8,power_law_1.01,0.06878592014312744
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,4,8,power_law_1.01,0.07238399982452393
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,4,8,power_law_1.01,0.07322879791259765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,4,8,power_law_1.01,0.07814015865325927
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,4,8,power_law_1.01,0.08133631706237793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,4,8,power_law_1.01,0.08536319732666016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,4,8,balanced,0.3258995056152344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,4,8,power_law_1.01,0.08070400238037109
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,4,8,power_law_1.01,0.08200832366943359
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,4,8,power_law_1.01,0.08136320114135742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,4,8,power_law_1.01,0.08561792373657226
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,4,8,power_law_1.01,0.09304320335388183
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,4,8,power_law_1.01,0.10441599845886232
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,4,8,power_law_1.01,0.1334425640106201
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,4,8,power_law_1.01,0.15373824119567872
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,4,8,power_law_1.01,0.22459007263183595
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,4,8,power_law_1.01,0.2905638313293457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,4,8,power_law_1.01,0.4241433715820312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,4,8,power_law_1.2,0.028395519256591794
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,4,8,power_law_1.01,0.5439846420288086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,4,8,power_law_1.2,0.03714432001113892
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,4,8,power_law_1.2,0.04124415874481201
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,4,8,power_law_1.01,0.6979558563232422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,4,8,power_law_1.2,0.04356224060058594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,4,8,power_law_1.2,0.04854144096374512
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,4,8,power_law_1.2,0.06616191864013672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,4,8,power_law_1.2,0.07011839866638184
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,4,8,power_law_1.2,0.0710323190689087
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,4,8,power_law_1.2,0.07410687923431396
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,4,8,power_law_1.01,1.01272705078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,4,8,power_law_1.2,0.07493504047393798
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,4,8,power_law_1.2,0.07863423824310303
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,4,8,power_law_1.2,0.08078335762023926
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,4,8,power_law_1.2,0.0846668815612793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,4,8,power_law_1.2,0.08305279731750488
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,4,8,power_law_1.2,0.08413311958312988
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,4,8,power_law_1.2,0.08826623916625977
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,4,8,power_law_1.2,0.09492223739624024
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,4,8,power_law_1.2,0.10907391548156738
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,4,8,power_law_1.01,0.07552256107330321
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,4,8,power_law_1.2,0.14433152198791505
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,4,8,power_law_1.2,0.17356544494628906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,4,8,power_law_1.2,0.2523404884338379
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,4,8,power_law_1.2,0.3089689636230469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,8,4,balanced,0.022076160907745362
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,4,8,power_law_1.2,0.47997055053710935
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,8,4,balanced,0.02923392057418823
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,2,16,power_law_1.2,1.4263923645019532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,8,4,balanced,0.046033921241760256
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,8,4,balanced,0.06296703815460206
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,8,4,balanced,0.10300671577453613
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,4,8,power_law_1.2,0.08255999565124512
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,8,4,balanced,0.10523520469665529
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,8,4,balanced,0.07690368175506593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,8,4,balanced,0.1064038372039795
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,8,4,balanced,0.10864128112792967
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,8,4,balanced,0.11415679931640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,8,4,balanced,0.1180185604095459
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,8,4,balanced,0.10580608367919922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,8,4,balanced,0.11013376235961914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,8,4,balanced,0.10711935997009277
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,8,4,balanced,0.1098739242553711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,8,4,balanced,0.11368703842163086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,8,4,balanced,0.12298496246337891
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,8,4,balanced,0.13090304374694822
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,8,4,balanced,0.18402048110961916
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,8,4,balanced,0.023476479053497316
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,8,4,balanced,0.20222591400146483
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,4,8,power_law_1.2,0.9551411437988282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,8,4,balanced,0.2836070442199707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,8,4,balanced,0.32469505310058594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,8,4,power_law_1.01,0.024479999542236328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,8,4,balanced,0.4839539337158203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,8,4,power_law_1.01,0.04037759780883789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,8,4,balanced,0.5841740798950196
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,8,4,power_law_1.01,0.041647357940673826
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,8,4,power_law_1.01,0.05244800090789795
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,8,4,power_law_1.01,0.09483776092529297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,8,4,balanced,0.8519923400878906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,8,4,power_law_1.01,0.09866239547729491
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,8,4,power_law_1.01,0.09877375602722169
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,8,4,power_law_1.01,0.09972991943359374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,8,4,power_law_1.01,0.10508159637451171
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,8,4,power_law_1.01,0.10988544464111327
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,8,4,power_law_1.01,0.10461695671081543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,8,4,power_law_1.01,0.10602751731872559
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,4,8,power_law_1.2,1.2884172058105468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,8,4,power_law_1.01,0.10681728363037109
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,8,4,power_law_1.01,0.11018495559692383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,8,4,power_law_1.01,0.12026495933532715
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,4,8,power_law_1.2,0.615412483215332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,8,4,power_law_1.01,0.1332876777648926
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,8,4,power_law_1.01,0.15281408309936523
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,8,4,power_law_1.01,0.18982784271240233
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,8,4,power_law_1.01,0.23633024215698245
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,8,4,power_law_1.01,0.30385408401489256
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,8,4,balanced,1.1122700500488283
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,8,4,power_law_1.01,0.40637950897216796
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,8,4,power_law_1.01,0.5634188842773438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,8,4,power_law_1.2,0.02422271966934204
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,8,4,power_law_1.01,0.6913190460205078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,8,4,power_law_1.2,0.037576959133148194
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,8,4,power_law_1.2,0.043692798614501954
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,8,4,power_law_1.01,0.0583948802947998
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,8,4,power_law_1.01,0.06993023872375488
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,8,4,power_law_1.2,0.047551999092102046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,8,4,power_law_1.01,0.9782502746582031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,8,4,power_law_1.2,0.09239680290222167
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,8,4,power_law_1.2,0.0947532844543457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,8,4,power_law_1.2,0.09828736305236815
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,8,4,power_law_1.2,0.10117504119873047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,8,4,power_law_1.2,0.11087743759155273
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,8,4,power_law_1.2,0.1043238353729248
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,8,4,power_law_1.2,0.10482687950134277
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,8,4,power_law_1.2,0.10708736419677733
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,8,4,power_law_1.2,0.11357567787170411
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,8,4,power_law_1.2,0.12121088027954101
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,8,4,power_law_1.2,0.13683327674865722
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,8,4,power_law_1.2,0.15491968154907226
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,8,4,power_law_1.2,0.20523263931274416
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,8,4,power_law_1.2,0.24368000030517578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,8,4,power_law_1.2,0.31031423568725586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,8,4,power_law_1.2,0.10839167594909668
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,8,4,power_law_1.01,1.239477767944336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,8,4,power_law_1.2,0.41846145629882814
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,16,2,balanced,0.021155838966369626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,16,2,balanced,0.02290560007095337
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,8,4,power_law_1.2,0.6204044723510742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,16,2,balanced,0.030655999183654786
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,16,2,balanced,0.048936958312988284
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,8,4,power_law_1.2,0.05876095771789551
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,8,4,power_law_1.2,0.0701900815963745
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,16,2,balanced,0.08739583969116212
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,16,2,balanced,0.09005824089050293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,16,2,balanced,0.09114368438720703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,16,2,balanced,0.10576383590698242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,16,2,balanced,0.10838656425476074
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,8,4,power_law_1.2,1.0847743988037108
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,8,4,power_law_1.2,0.7100876617431641
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,16,2,balanced,0.10917759895324708
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,16,2,balanced,0.11147007942199708
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,16,2,balanced,0.11402239799499511
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,16,2,balanced,0.1116198444366455
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,16,2,balanced,0.11361920356750488
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,16,2,balanced,0.1158566379547119
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,16,2,balanced,0.12450559616088867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,16,2,balanced,0.13104512214660644
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,16,2,balanced,0.19044095993041993
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,16,2,balanced,0.20397567749023438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,16,2,balanced,0.3061939239501953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,16,2,balanced,0.3375948715209961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,16,2,balanced,0.07115776062011718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,16,2,balanced,0.509093132019043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,16,2,power_law_1.01,0.022364161014556884
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,16,2,balanced,0.10629887580871582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,16,2,balanced,0.617573127746582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,16,2,power_law_1.01,0.040538878440856935
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,16,2,power_law_1.01,0.05815552234649658
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,16,2,power_law_1.01,0.08038784027099609
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,16,2,power_law_1.01,0.07449344158172608
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,16,2,balanced,0.9133977508544922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,16,2,power_law_1.01,0.08432255744934082
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,16,2,balanced,1.1903616333007814
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,16,2,power_law_1.01,0.03250175952911377
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,16,2,power_law_1.01,0.0498419189453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,16,2,power_law_1.01,0.10081024169921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,16,2,power_law_1.01,0.09851776123046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,16,2,power_law_1.01,0.1025011157989502
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,16,2,power_law_1.01,0.10356351852416992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,16,2,power_law_1.01,0.11569664001464844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,16,2,power_law_1.01,0.12323455810546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,16,2,power_law_1.01,0.13850367546081543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,16,2,power_law_1.01,0.10709376335144043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,16,2,power_law_1.01,0.15695487976074218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,16,2,power_law_1.01,0.18923648834228515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,16,2,power_law_1.01,0.24568832397460935
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,16,2,power_law_1.01,0.30857343673706056
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,16,2,power_law_1.01,0.11158656120300292
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,16,2,power_law_1.01,0.3914547348022461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,16,2,power_law_1.2,0.02237567901611328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,16,2,power_law_1.01,0.5436326217651367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,8,4,power_law_1.2,1.4226841735839844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,16,2,power_law_1.01,0.7047872161865234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,16,2,power_law_1.2,0.04058879852294922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,16,2,power_law_1.01,0.11229439735412598
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,16,2,power_law_1.2,0.05562111854553222
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,16,2,power_law_1.2,0.07929599761962891
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,16,2,power_law_1.01,1.0418956756591797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,16,2,power_law_1.2,0.09994624137878419
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,16,2,power_law_1.01,1.2992588806152345
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,16,2,power_law_1.2,0.10478207588195801
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,16,2,power_law_1.2,0.10483327865600586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,16,2,power_law_1.2,0.10761728286743164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,16,2,power_law_1.2,0.11259391784667969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,16,2,power_law_1.2,0.11432959556579589
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,16,2,power_law_1.2,0.07417856216430664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,16,2,power_law_1.2,0.03297663927078247
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,16,2,power_law_1.2,0.11947648048400879
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,16,2,power_law_1.2,0.12666879653930666
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,16,2,power_law_1.2,0.1416985607147217
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,16,2,power_law_1.2,0.1604172706604004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,16,2,power_law_1.2,0.22159103393554686
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,16,2,power_law_1.2,0.2516787147521973
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,16,2,power_law_1.2,0.3141862487792969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,16,2,power_law_1.2,0.4127641677856445
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,16,2,power_law_1.2,0.5772159957885743
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,16,2,power_law_1.2,0.7380902099609374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,16,2,power_law_1.2,0.10036735534667969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,32,1,balanced,0.03265791893005371
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,16,2,power_law_1.2,0.08227968215942383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,32,1,balanced,0.06646272182464599
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,16,2,power_law_1.2,1.1048896026611328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,32,1,balanced,0.09952128410339356
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,32,1,balanced,0.12498047828674316
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,32,1,balanced,0.12668416023254395
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,32,1,balanced,0.12707903861999512
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,32,1,balanced,0.1272435188293457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,32,1,balanced,0.12876352310180664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,32,1,balanced,0.13172672271728517
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,16,2,power_law_1.2,0.046385917663574215
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,32,1,balanced,0.13369152069091797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,32,1,balanced,0.13491583824157716
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,32,1,balanced,0.13803775787353517
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,32,1,balanced,0.13592512130737305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,32,1,balanced,0.13718015670776368
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,32,1,balanced,0.13981375694274903
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,32,1,balanced,0.14687616348266602
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,32,1,balanced,0.15298879623413086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,32,1,balanced,0.2245580863952637
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,32,1,balanced,0.027402238845825194
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,32,1,balanced,0.2338630485534668
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,32,1,balanced,0.34566078186035154
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,32,1,balanced,0.3694521713256836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,32,1,balanced,0.04620160102844238
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,32,1,power_law_1.01,0.027641599178314206
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,32,1,balanced,0.5673612976074219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,32,1,power_law_1.01,0.03236160039901733
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,32,1,power_law_1.01,0.049869441986083986
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,32,1,balanced,0.6787872314453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,32,1,power_law_1.01,0.06301887989044189
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,32,1,power_law_1.01,0.07636223793029785
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,32,1,power_law_1.01,0.10514944076538087
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,32,1,power_law_1.01,0.10867839813232423
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,32,1,power_law_1.01,0.11397312164306642
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,32,1,balanced,1.0167660522460937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,32,1,power_law_1.01,0.1209932804107666
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,32,1,power_law_1.01,0.12691455841064453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,32,1,power_law_1.01,0.12900416374206541
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,32,1,power_law_1.01,0.13295104026794433
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,32,1,power_law_1.01,0.13523839950561523
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,32,1,power_law_1.01,0.13888447761535644
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,32,1,power_law_1.01,0.14656319618225097
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,32,1,power_law_1.01,0.16436927795410156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,32,1,power_law_1.01,0.18070144653320314
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,32,1,power_law_1.01,0.2131078338623047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,32,1,power_law_1.01,0.1172480010986328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,32,1,power_law_1.01,0.27393728256225586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,32,1,power_law_1.01,0.12968704223632813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,32,1,power_law_1.01,0.3410636901855469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,32,1,power_law_1.01,0.41350654602050785
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,32,1,balanced,1.3385548400878906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,32,1,power_law_1.2,0.027733759880065916
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,32,1,power_law_1.01,0.5791820907592774
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,32,1,power_law_1.2,0.03834304094314575
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,32,1,power_law_1.2,0.049915518760681156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,32,1,power_law_1.2,0.05981823921203613
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,32,1,power_law_1.01,0.7431839752197266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,32,1,power_law_1.2,0.10343232154846191
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,32,1,power_law_1.2,0.10950783729553222
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,32,1,power_law_1.2,0.11476415634155272
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,16,2,power_law_1.2,1.3281228637695313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,32,1,power_law_1.2,0.12410431861877443
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,32,1,power_law_1.2,0.13170432090759276
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,32,1,power_law_1.2,0.12901311874389648
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,32,1,power_law_1.2,0.1320518398284912
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,32,1,power_law_1.2,0.13717568397521973
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,32,1,power_law_1.2,0.13842111587524414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,32,1,power_law_1.01,1.3967546081542968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,32,1,power_law_1.2,0.143055362701416
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,32,1,power_law_1.2,0.15025728225708007
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,32,1,power_law_1.2,0.07296512126922608
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,32,1,power_law_1.2,0.18176767349243164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,32,1,power_law_1.2,0.21674495697021484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,32,1,power_law_1.2,0.11854592323303223
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,32,1,power_law_1.01,1.0532275390624999
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,32,1,power_law_1.2,0.34709312438964846
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,32,1,power_law_1.2,0.4131590270996094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,1,32,balanced,0.05634175777435303
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,1,32,balanced,0.07518464088439941
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,32,1,power_law_1.2,0.16803647994995116
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,1,32,balanced,0.11539199829101562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,1,32,balanced,0.2376947212219238
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,32,1,power_law_1.2,0.2769171142578125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,1,32,balanced,0.30087039947509764
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,32,1,power_law_1.2,0.5799596786499024
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,1,32,balanced,0.4341939163208008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,1,32,balanced,0.4428492736816406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,1,32,balanced,0.4316211318969726
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,1,32,balanced,0.43724800109863277
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,1,32,balanced,0.4493683242797852
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,1,32,balanced,0.4520793533325195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,1,32,balanced,0.4576819229125976
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,1,32,balanced,0.4734425735473633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,32,1,power_law_1.2,0.7486278533935546
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,1,32,balanced,0.482872314453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,1,32,balanced,0.5005721664428711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,1,32,balanced,0.4964697647094726
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,1,32,balanced,0.520081901550293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,1,32,balanced,0.573548812866211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,1,32,balanced,0.6244736099243164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,1,32,balanced,0.7690585327148438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,1,32,balanced,0.8600537872314453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,1,32,balanced,1.159557113647461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,1,32,balanced,0.03401087999343872
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,1,32,balanced,1.3606118774414062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,1,32,balanced,0.04886528015136719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,1,32,balanced,0.0656217622756958
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,1,32,balanced,0.11624704360961916
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,1,32,balanced,0.15885184288024903
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,1,32,balanced,0.17319551467895508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,1,32,balanced,0.19220352172851562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,1,32,balanced,2.0895143127441407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,1,32,balanced,0.18337791442871093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,1,32,balanced,0.18460416793823242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,1,32,balanced,0.19632768630981445
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,1,32,balanced,0.200262393951416
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,1,32,balanced,0.20478464126586915
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,1,32,balanced,2.685103454589844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,32,1,power_law_1.2,1.070666275024414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,1,32,balanced,0.21936128616333006
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,1,32,balanced,0.23133440017700196
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,1,32,balanced,0.249800968170166
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,1,32,balanced,0.240053768157959
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,1,32,balanced,0.25701759338378904
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,1,32,balanced,0.3019033622741699
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,1,32,balanced,0.34142208099365234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,1,32,balanced,0.42115329742431645
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,1,32,balanced,0.502487030029297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,1,32,balanced,0.7225676727294922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,1,32,balanced,4.062009582519531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,1,32,balanced,0.8948710632324218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,1,32,power_law_1.01,0.2913484764099121
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,1,32,balanced,1.2709632110595703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,1,32,power_law_1.01,0.29306495666503907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,1,32,balanced,1.5925875854492186
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,1,32,power_law_1.01,0.29407104492187497
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,1,32,power_law_1.01,0.29667327880859373
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,1,32,balanced,5.312771606445312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,32,1,power_law_1.2,1.4110368347167968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,1,32,power_law_1.01,0.39700096130371093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,1,32,power_law_1.01,0.39709823608398437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,1,32,power_law_1.01,0.4006399917602539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,1,32,power_law_1.01,0.4061747360229492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,1,32,balanced,2.389171142578125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,1,32,power_law_1.01,0.43089920043945307
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,1,32,power_law_1.01,0.431146240234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,1,32,power_law_1.01,0.4504191970825195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,1,32,power_law_1.01,0.4697216033935547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,1,32,power_law_1.01,0.2763545608520508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,1,32,power_law_1.01,0.5026176071166992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,1,32,power_law_1.01,0.5346854400634766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,1,32,power_law_1.01,0.5386751937866211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,1,32,power_law_1.01,0.5573900985717773
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,1,32,balanced,3.12482177734375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,1,32,power_law_1.01,0.6313983917236328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,1,32,power_law_1.01,0.7153561401367188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,1,32,power_law_1.01,0.9208767700195313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,1,32,power_law_1.01,1.2095577239990234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,1,32,power_law_1.01,0.1509017562866211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,1,32,power_law_1.01,0.15183232307434083
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,1,32,power_law_1.01,0.15104512214660643
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,1,32,power_law_1.01,1.6154368591308592
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,1,32,power_law_1.01,0.15369600296020508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,1,32,power_law_1.01,0.14648447990417482
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,1,32,power_law_1.01,0.1661631965637207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,1,32,power_law_1.01,2.196408386230469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,1,32,power_law_1.01,0.17255807876586915
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,1,32,power_law_1.01,0.17393791198730468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,1,32,power_law_1.01,0.17740543365478517
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,1,32,power_law_1.01,0.17964544296264648
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,1,32,power_law_1.01,0.19030399322509767
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,1,32,power_law_1.01,0.2034931182861328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,1,32,power_law_1.01,0.21582208633422853
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,1,32,power_law_1.01,0.23694080352783203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,1,32,power_law_1.01,3.080814208984375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,1,32,power_law_1.01,0.24487424850463868
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,1,32,power_law_1.01,0.24388351440429687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,1,32,power_law_1.01,0.26905855178833005
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,1,32,power_law_1.01,0.3368473434448242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,1,32,power_law_1.01,0.37296768188476564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,1,32,power_law_1.01,0.5098406219482422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,1,32,power_law_1.01,0.6510387420654297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,1,32,power_law_1.01,4.446155395507812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,1,32,power_law_1.01,0.8425177764892577
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,1,32,power_law_1.01,1.1338380432128905
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,1,32,power_law_1.2,0.2946739196777344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,1,32,power_law_1.01,1.6376626586914065
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,1,32,power_law_1.2,0.29340032577514646
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,1,32,power_law_1.01,6.267097778320313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,1,32,power_law_1.2,0.29475967407226566
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,1,32,power_law_1.2,0.2976358413696289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,1,32,power_law_1.01,2.289311981201172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,1,32,power_law_1.2,0.30210687637329103
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,1,32,power_law_1.2,0.3909414291381836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,1,32,power_law_1.2,0.38015361785888674
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,1,32,power_law_1.2,0.38147838592529293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,1,32,power_law_1.2,0.4123199844360352
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,1,32,power_law_1.2,0.41109886169433596
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,1,32,power_law_1.2,0.447059211730957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,1,32,power_law_1.2,0.47347839355468746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,1,32,power_law_1.2,0.48876800537109377
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,1,32,power_law_1.01,3.274569091796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,1,32,power_law_1.2,0.5320012664794922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,1,32,power_law_1.2,0.5311974334716797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,1,32,power_law_1.01,8.3229248046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,1,32,power_law_1.2,0.5427123260498047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,1,32,power_law_1.2,0.5738291168212891
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,1,32,power_law_1.2,0.6566413116455079
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,1,32,power_law_1.2,0.7752588653564454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,1,32,power_law_1.01,4.271734924316407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,1,32,power_law_1.2,0.9659468841552734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,1,32,power_law_1.2,0.1507686424255371
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,1,32,power_law_1.2,0.1507571220397949
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,1,32,power_law_1.2,2.0109645080566407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,1,32,power_law_1.2,0.15182592391967772
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,1,32,power_law_1.2,1.3555711364746093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,1,32,power_law_1.2,2.4783795166015627
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,1,32,power_law_1.2,0.1544051170349121
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,1,32,power_law_1.2,0.14705151557922364
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,1,32,power_law_1.2,0.1665318489074707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,1,32,power_law_1.2,0.16777088165283202
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,1,32,power_law_1.2,0.17828992843627928
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,1,32,power_law_1.2,0.1812940788269043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,1,32,power_law_1.2,0.1855731201171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,1,32,power_law_1.2,0.19879936218261718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,1,32,power_law_1.2,0.20543487548828124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,1,32,power_law_1.2,0.22117887496948244
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,1,32,power_law_1.2,0.24586111068725586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,1,32,power_law_1.2,0.2472768020629883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,1,32,power_law_1.2,0.24434303283691405
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,1,32,power_law_1.2,4.017632141113281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,1,32,power_law_1.2,0.29095680236816407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,1,32,power_law_1.2,0.35078399658203124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,1,32,power_law_1.2,0.41407104492187496
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,1,32,power_law_1.2,0.531256332397461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,1,32,power_law_1.2,0.6907750701904296
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,1,32,power_law_1.2,0.9897714996337891
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,1,32,power_law_1.2,5.6032385253906245
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,1,32,power_law_1.2,1.2592499542236328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,2,16,balanced,0.054319357872009276
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,2,16,balanced,0.06607359886169434
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,2,16,balanced,0.08650495529174804
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,2,16,balanced,0.16153343200683595
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,1,32,power_law_1.2,1.7988748168945314
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,2,16,balanced,0.2651840019226074
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,2,16,balanced,0.2900351905822754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,2,16,balanced,0.4182271957397461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,2,16,balanced,0.42047489166259766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,1,32,power_law_1.2,2.5268992614746093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,2,16,balanced,0.4309862518310547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,2,16,balanced,0.4347225570678711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,2,16,balanced,0.4457088088989257
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,2,16,balanced,0.46655105590820317
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,2,16,balanced,0.4536665725708008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,2,16,balanced,0.4847526550292969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,2,16,balanced,0.4780774307250977
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,2,16,balanced,0.4951859283447265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,2,16,balanced,0.5275059127807618
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,1,32,power_law_1.2,8.464066772460937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,2,16,balanced,0.5951372909545898
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,2,16,balanced,0.6535167694091797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,1,32,power_law_1.2,3.943376770019531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,2,16,balanced,0.7889485168457031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,2,16,balanced,0.8695513916015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,2,16,balanced,1.2326195526123047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,2,16,balanced,1.4646438598632812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,2,16,balanced,0.03141247987747192
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,2,16,balanced,0.041486082077026365
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,2,16,balanced,0.057541117668151856
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,2,16,balanced,0.09700096130371094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,2,16,balanced,0.1474048042297363
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,2,16,balanced,2.1990988159179685
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,2,16,balanced,0.16958208084106446
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,2,16,balanced,0.18544767379760743
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,2,16,balanced,0.18817535400390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,2,16,balanced,0.19792768478393555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,2,16,balanced,0.2005414390563965
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,2,16,balanced,0.20566911697387696
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,2,16,balanced,2.8297100830078126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,2,16,balanced,0.2238847923278809
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,2,16,balanced,0.2388096046447754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,2,16,balanced,0.25304704666137695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,2,16,balanced,0.2562380790710449
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,2,16,balanced,0.2743628883361816
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,2,16,balanced,0.3233638381958008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,2,16,balanced,0.2373017692565918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,2,16,balanced,0.374257926940918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,1,32,power_law_1.2,5.13712890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,2,16,balanced,4.317004699707032
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,2,16,balanced,0.46555137634277344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,2,16,balanced,0.560898551940918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,1,32,power_law_1.2,10.90325927734375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,2,16,balanced,0.8151859283447266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,2,16,balanced,0.996659164428711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,2,16,power_law_1.01,0.23161088943481448
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,2,16,balanced,5.636063842773437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,2,16,balanced,1.4503424072265625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,2,16,power_law_1.01,0.26860416412353516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,2,16,power_law_1.01,0.27089279174804687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,2,16,power_law_1.01,0.2722112083435059
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,2,16,power_law_1.01,0.2768819236755371
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,2,16,balanced,1.822923583984375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,2,16,power_law_1.01,0.2769804763793945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,2,16,power_law_1.01,0.38974720001220703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,2,16,power_law_1.01,0.3798400115966797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,2,16,power_law_1.01,0.39727870941162113
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,2,16,power_law_1.01,0.4127360153198242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,2,16,power_law_1.01,0.41849216461181643
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,2,16,power_law_1.01,0.45294464111328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,2,16,power_law_1.01,0.4819878387451172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,2,16,power_law_1.01,0.47935359954833984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,2,16,balanced,2.7475250244140623
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,2,16,power_law_1.01,0.4814144134521484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,2,16,power_law_1.01,0.5162944030761719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,2,16,power_law_1.01,0.5732352066040038
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,2,16,power_law_1.01,0.6687449645996094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,2,16,balanced,3.5628826904296873
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,2,16,power_law_1.01,0.4472243118286133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,2,16,power_law_1.01,0.8174323272705077
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,2,16,power_law_1.01,1.0603008270263672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,2,16,power_law_1.01,0.11483648300170898
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,2,16,power_law_1.01,1.4435877990722656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,2,16,power_law_1.01,0.15432191848754884
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,2,16,power_law_1.01,0.15274111747741698
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,2,16,power_law_1.01,1.8165504455566406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,2,16,power_law_1.01,0.15647359848022463
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,2,16,power_law_1.01,0.15880576133728028
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,2,16,power_law_1.01,0.15760128021240233
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,2,16,power_law_1.01,0.17927040100097655
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,2,16,power_law_1.01,0.18435327529907225
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,2,16,power_law_1.01,0.18810880661010743
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,2,16,power_law_1.01,0.19029632568359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,2,16,power_law_1.01,0.1920140838623047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,2,16,power_law_1.01,0.20744447708129882
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,2,16,power_law_1.01,2.64371337890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,2,16,power_law_1.01,0.2418035125732422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,2,16,power_law_1.01,0.22827264785766604
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,2,16,power_law_1.01,0.23918848037719725
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,2,16,power_law_1.01,0.26546688079833985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,2,16,power_law_1.01,0.32165374755859377
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,2,16,power_law_1.01,0.359185905456543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,2,16,power_law_1.01,3.532670593261719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,2,16,power_law_1.01,0.21728384017944338
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,2,16,power_law_1.01,0.6001433563232422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,2,16,power_law_1.01,0.783114242553711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,2,16,power_law_1.01,1.0396173095703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,2,16,power_law_1.01,0.46556545257568355
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,2,16,power_law_1.2,0.23170047760009765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,2,16,power_law_1.01,5.0052926635742185
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,2,16,power_law_1.01,1.5200743103027343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,2,16,power_law_1.2,0.2628659248352051
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,2,16,power_law_1.2,0.2665279960632324
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,2,16,power_law_1.2,0.27543935775756834
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,2,16,power_law_1.01,1.980482635498047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,2,16,power_law_1.2,0.27908735275268554
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,2,16,power_law_1.2,0.2762303924560547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,2,16,power_law_1.2,0.38638206481933596
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,2,16,power_law_1.2,0.3825600051879883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,2,16,power_law_1.01,6.31390869140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,2,16,power_law_1.2,0.4128153610229492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,2,16,power_law_1.2,0.4052799987792969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,2,16,power_law_1.2,0.42951934814453124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,2,16,power_law_1.01,2.7825537109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,2,16,power_law_1.2,0.44958847045898437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,2,16,power_law_1.2,0.4956390380859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,2,16,power_law_1.2,0.46057983398437496
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,2,16,power_law_1.2,0.4971072006225586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,2,16,power_law_1.2,0.6148083114624023
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,2,16,power_law_1.2,0.6565209960937499
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,2,16,power_law_1.2,0.4432371139526367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,2,16,power_law_1.2,0.8547698974609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,2,16,power_law_1.2,1.0767167663574218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,2,16,power_law_1.2,0.4704806518554688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,2,16,power_law_1.2,0.11452544212341309
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,2,16,power_law_1.2,1.53725830078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,2,16,power_law_1.2,0.14954112052917481
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,2,16,power_law_1.01,3.830025024414063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,2,16,power_law_1.2,0.15393407821655275
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,2,16,power_law_1.2,1.8787263488769532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,2,16,power_law_1.2,0.15538175582885744
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,2,16,power_law_1.2,0.17143680572509765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,2,16,power_law_1.2,0.1581375980377197
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,2,16,power_law_1.2,0.18614784240722657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,2,16,power_law_1.2,0.17911552429199218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,2,16,power_law_1.2,0.18831104278564453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,2,16,power_law_1.2,0.1906892776489258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,2,16,power_law_1.2,2.810624084472656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,2,16,power_law_1.2,0.19558399200439452
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,2,16,power_law_1.2,0.20617855072021482
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,2,16,power_law_1.2,0.225030403137207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,2,16,power_law_1.2,0.24730752944946288
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,2,16,power_law_1.2,0.23514879226684568
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,2,16,power_law_1.2,0.23040895462036132
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,2,16,power_law_1.2,0.2669785690307617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,2,16,power_law_1.2,0.3256332778930664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,2,16,power_law_1.2,0.3853401565551758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,2,16,power_law_1.2,0.5043123245239258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,2,16,power_law_1.2,0.6318835067749023
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,2,16,power_law_1.2,0.829244155883789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,2,16,power_law_1.2,1.0539929962158203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,2,16,power_law_1.2,4.2179980468750005
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,4,8,balanced,0.04592127799987793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,4,8,balanced,0.05958528041839599
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,4,8,balanced,0.1274304008483887
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,2,16,power_law_1.2,1.4938983154296874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,4,8,balanced,0.24553728103637695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,4,8,balanced,0.2834252738952637
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,2,16,power_law_1.2,2.0537767028808593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,4,8,balanced,0.2971673583984375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,4,8,balanced,0.3027609634399414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,4,8,balanced,0.43572353363037114
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,4,8,balanced,0.4357747268676758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,4,8,balanced,0.43960193634033207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,4,8,balanced,0.4533478546142578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,4,8,balanced,0.4578879928588867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,4,8,balanced,0.07853824138641358
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,2,16,power_law_1.2,3.064049987792969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,2,16,power_law_1.2,6.051201171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,4,8,balanced,0.48150398254394533
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,4,8,balanced,0.47153152465820314
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,4,8,balanced,0.4833804702758789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,2,16,power_law_1.2,7.992279052734375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,4,8,balanced,0.5115033721923827
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,4,8,balanced,0.5724812698364258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,4,8,balanced,0.6243199920654297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,4,8,balanced,0.7414796447753906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,4,8,balanced,0.8386406707763673
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,4,8,balanced,0.026686720848083496
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,4,8,balanced,1.1758758544921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,4,8,balanced,0.03904000043869019
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,4,8,balanced,0.05312384128570556
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,4,8,balanced,1.3868850708007812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,4,8,balanced,0.09179136276245117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,4,8,balanced,0.12553343772888184
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,4,8,balanced,0.17255039215087892
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,4,8,balanced,0.17700735092163086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,4,8,balanced,0.2048716735839844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,4,8,balanced,2.078513946533203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,4,8,balanced,0.20676095962524413
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,4,8,balanced,0.21284223556518556
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,4,8,balanced,0.22952320098876955
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,4,8,balanced,0.23917055130004883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,4,8,balanced,0.25533824920654297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,4,8,balanced,2.6389553833007815
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,4,8,balanced,0.23768320083618164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,4,8,balanced,0.2546777534484863
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,4,8,balanced,0.16453376770019532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,4,8,balanced,0.2705331230163574
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,4,8,balanced,0.3121254348754883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,4,8,balanced,0.35796993255615234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,4,8,balanced,0.4419353485107422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,4,8,balanced,0.5185868835449219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,4,8,balanced,0.7721663665771484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,4,8,balanced,3.9145703124999995
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,4,8,balanced,0.9377613067626953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,4,8,power_law_1.01,0.1157811164855957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,4,8,balanced,1.366033935546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,4,8,power_law_1.01,0.23491712570190432
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,4,8,balanced,5.072749938964844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,4,8,power_law_1.01,0.23745023727416995
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,4,8,power_law_1.01,0.25111295700073244
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,4,8,power_law_1.01,0.2815475273132324
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,4,8,power_law_1.01,0.27451263427734374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,4,8,balanced,2.559944915771484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,2,16,power_law_1.2,4.085726623535156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,4,8,power_law_1.01,0.39771648406982424
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,4,8,power_law_1.01,0.4217305755615234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,4,8,balanced,1.6918348693847656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,4,8,power_law_1.01,0.4131468963623047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,4,8,power_law_1.01,0.4411123275756836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,4,8,power_law_1.01,0.18051456451416015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,4,8,power_law_1.01,0.4732505416870117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,4,8,power_law_1.01,0.45564800262451166
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,4,8,balanced,3.322496032714844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,4,8,power_law_1.01,0.4715135955810547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,4,8,power_law_1.01,0.49839870452880863
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,4,8,power_law_1.01,0.553807373046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,4,8,power_law_1.01,0.2774297523498535
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,4,8,power_law_1.01,0.6257126235961914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,4,8,power_law_1.01,0.7627967834472656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,4,8,power_law_1.01,0.4398668670654297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,4,8,power_law_1.01,0.9790092468261719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,4,8,power_law_1.01,1.2620851135253905
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,4,8,power_law_1.01,0.070348801612854
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,4,8,power_law_1.01,1.5687385559082032
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,4,8,power_law_1.01,0.10341631889343261
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,4,8,power_law_1.01,0.11688575744628907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,4,8,power_law_1.01,0.12597503662109374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,4,8,power_law_1.01,0.14961024284362795
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,4,8,power_law_1.01,0.15366144180297853
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,4,8,power_law_1.01,2.2439424133300783
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,4,8,power_law_1.01,0.16436096191406252
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,4,8,power_law_1.01,0.1915443229675293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,4,8,power_law_1.01,0.19537408828735353
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,4,8,power_law_1.01,0.19651968002319337
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,4,8,power_law_1.01,0.21693183898925783
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,4,8,power_law_1.01,2.995605773925781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,4,8,power_law_1.01,0.22563711166381836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,4,8,power_law_1.01,0.2448217582702637
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,4,8,power_law_1.01,0.22608768463134768
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,4,8,power_law_1.01,0.2417215919494629
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,4,8,power_law_1.01,0.2651161575317383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,4,8,power_law_1.01,0.31848064422607425
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,4,8,power_law_1.01,0.3686796951293945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,4,8,power_law_1.01,0.47016448974609376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,4,8,power_law_1.01,0.5924607849121093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,4,8,power_law_1.01,0.7718643188476563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,4,8,power_law_1.01,0.9946380615234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,4,8,power_law_1.2,0.10871808052062988
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,4,8,power_law_1.2,0.13589759826660158
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,4,8,power_law_1.01,4.345063781738281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,4,8,power_law_1.01,1.8350015258789063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,4,8,power_law_1.2,0.19349376678466795
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,4,8,power_law_1.01,5.885203247070312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,4,8,power_law_1.2,0.22279935836791992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,4,8,power_law_1.01,0.16234880447387695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,4,8,power_law_1.01,1.3949964904785157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,4,8,power_law_1.2,0.2690803146362305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,4,8,power_law_1.2,0.284901123046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,4,8,power_law_1.2,0.2783935928344726
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,4,8,power_law_1.2,0.39500926971435546
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,4,8,power_law_1.01,2.730310363769531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,4,8,power_law_1.2,0.4076300811767578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,4,8,power_law_1.2,0.4398886489868164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,4,8,power_law_1.2,0.45197185516357424
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,4,8,power_law_1.2,0.478869743347168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,4,8,power_law_1.2,0.4628185653686524
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,4,8,power_law_1.2,0.4699558258056641
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,4,8,power_law_1.2,0.49154304504394536
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,4,8,power_law_1.2,0.5565875244140626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,4,8,power_law_1.01,3.531555786132812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,4,8,power_law_1.2,0.42986110687255863
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,4,8,power_law_1.2,0.6381811141967774
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,4,8,power_law_1.2,0.7854541015625001
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,4,8,power_law_1.2,0.2728691291809082
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,4,8,power_law_1.2,0.948569564819336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,4,8,power_law_1.2,0.07004799842834472
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,4,8,power_law_1.2,0.09065983772277833
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,4,8,power_law_1.2,1.3262937927246095
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,4,8,power_law_1.2,0.10467583656311034
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,4,8,power_law_1.2,0.12637439727783203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,4,8,power_law_1.2,1.7375820922851564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,4,8,power_law_1.2,0.14963839530944825
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,4,8,power_law_1.2,0.15169024467468262
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,4,8,power_law_1.2,0.16128896713256835
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,4,8,power_law_1.2,0.15768320083618165
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,4,8,power_law_1.2,0.1884979248046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,4,8,power_law_1.2,0.2022195243835449
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,4,8,power_law_1.2,0.21775999069213867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,4,8,power_law_1.2,0.22824703216552736
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,4,8,power_law_1.2,0.2521075248718262
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,4,8,power_law_1.2,0.233941764831543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,4,8,power_law_1.2,0.25264896392822267
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,4,8,power_law_1.2,0.19137535095214844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,4,8,power_law_1.2,0.2897548866271973
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,4,8,power_law_1.2,0.325164794921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,4,8,power_law_1.2,0.3784422302246094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,4,8,power_law_1.2,2.4222706604003905
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,4,8,power_law_1.2,0.4889535903930664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,4,8,power_law_1.2,0.6053311920166016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,4,8,power_law_1.2,0.8066726684570312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,4,8,power_law_1.2,3.2862170410156253
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,8,4,balanced,0.042260479927062986
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,8,4,balanced,0.05349376201629639
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,4,8,power_law_1.2,1.43251708984375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,8,4,balanced,0.0728115177154541
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,8,4,balanced,0.11879039764404298
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,8,4,balanced,0.24338943481445313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,8,4,balanced,0.28611455917358397
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,8,4,balanced,0.3004748725891113
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,4,8,power_law_1.2,1.0176512145996095
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,4,8,power_law_1.2,1.9507852172851563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,4,8,power_law_1.2,4.753897094726563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,8,4,balanced,0.30373376846313477
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,8,4,balanced,0.31537919998168945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,8,4,balanced,0.3197273635864258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,8,4,balanced,0.3268454360961914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,8,4,balanced,0.4566336059570313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,8,4,balanced,0.44566528320312504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,8,4,balanced,0.4733465576171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,8,4,balanced,0.4591974258422852
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,8,4,balanced,0.47403007507324213
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,4,8,power_law_1.2,2.8484442138671873
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,8,4,balanced,0.49781631469726567
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,8,4,balanced,0.5482867050170899
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,8,4,balanced,0.5985599899291992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,8,4,balanced,0.7506841278076172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,4,8,power_law_1.2,6.33172607421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,8,4,balanced,0.850613784790039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,8,4,balanced,1.2092582702636718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,8,4,balanced,0.02823807954788208
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,8,4,balanced,1.4196954345703126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,8,4,balanced,0.03821952104568481
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,8,4,balanced,0.05447679996490479
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,8,4,balanced,0.09150464057922363
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,8,4,balanced,0.14297216415405273
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,8,4,balanced,2.1401625061035157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,8,4,balanced,0.18218751907348635
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,8,4,balanced,0.19475967407226563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,8,4,balanced,0.2044198417663574
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,8,4,balanced,0.2067865562438965
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,8,4,balanced,0.21178623199462893
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,8,4,balanced,2.712672119140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,8,4,balanced,0.24970495223999026
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,8,4,balanced,0.25726720809936526
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,8,4,balanced,0.2772863960266113
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,8,4,balanced,0.25854848861694335
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,8,4,balanced,0.19227392196655274
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,8,4,balanced,0.2908185577392578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,8,4,balanced,0.3297062301635742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,8,4,balanced,0.3641523361206055
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,8,4,balanced,0.43786880493164065
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,8,4,balanced,3.9960818481445317
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,8,4,balanced,0.5097177505493165
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,8,4,balanced,0.2672217559814453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,8,4,balanced,0.7860364532470703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,8,4,balanced,0.9310758209228516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,8,4,power_law_1.01,0.06904831886291504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,8,4,power_law_1.01,0.10666751861572266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,8,4,power_law_1.01,0.1105228805541992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,8,4,balanced,1.3645785522460936
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,8,4,balanced,5.150951538085938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,8,4,power_law_1.01,0.1454528045654297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,8,4,power_law_1.01,0.18616064071655275
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,8,4,balanced,1.6715046691894533
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,8,4,power_law_1.01,0.26984064102172856
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,8,4,power_law_1.01,0.27057279586791994
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,4,8,power_law_1.2,3.7403186035156253
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,8,4,power_law_1.01,0.2849663925170899
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,8,4,power_law_1.01,0.29690368652343746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,8,4,power_law_1.01,0.30915327072143556
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,8,4,balanced,2.5203878784179685
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,8,4,power_law_1.01,0.4429350280761719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,8,4,power_law_1.01,0.45414142608642577
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,8,4,power_law_1.01,0.2675289535522461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,8,4,power_law_1.01,0.47745025634765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,8,4,power_law_1.01,0.5385420989990235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,8,4,balanced,3.2416946411132814
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,8,4,power_law_1.01,0.6084492874145508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,8,4,power_law_1.01,0.7455423736572266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,8,4,power_law_1.01,0.8979827117919921
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,8,4,power_law_1.01,0.42864639282226563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,8,4,power_law_1.01,0.43948417663574213
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,8,4,power_law_1.01,1.198199005126953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,8,4,power_law_1.01,0.45763454437255857
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,8,4,power_law_1.01,0.04895743846893311
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,8,4,power_law_1.01,1.5533439636230468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,8,4,power_law_1.01,0.07551104068756104
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,8,4,power_law_1.01,0.1074188804626465
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,8,4,power_law_1.01,0.11928959846496583
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,8,4,power_law_1.01,0.1672166442871094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,8,4,power_law_1.01,2.1905255126953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,8,4,power_law_1.01,0.17620351791381836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,8,4,power_law_1.01,0.17527807235717774
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,8,4,power_law_1.01,0.1861836814880371
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,8,4,power_law_1.01,0.18661376953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,8,4,power_law_1.01,0.19307775497436525
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,8,4,power_law_1.01,0.23498367309570312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,8,4,power_law_1.01,0.08154623985290528
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,8,4,power_law_1.01,0.2444428825378418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,8,4,power_law_1.01,2.925420837402344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,8,4,power_law_1.01,0.246615047454834
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,8,4,power_law_1.01,0.2552000045776367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,8,4,power_law_1.01,0.2788902473449707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,8,4,power_law_1.01,0.33800193786621097
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,8,4,power_law_1.01,0.3879167938232422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,8,4,power_law_1.01,0.47327743530273436
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,8,4,power_law_1.01,0.5750336074829102
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,8,4,power_law_1.01,0.7697856140136718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,8,4,power_law_1.01,0.2613542366027832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,8,4,power_law_1.01,0.9740480041503906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,8,4,power_law_1.2,0.06902656078338623
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,8,4,power_law_1.01,4.275701904296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,8,4,power_law_1.2,0.08758655548095703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,8,4,power_law_1.2,0.11230463981628418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,8,4,power_law_1.01,1.3786636352539063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,8,4,power_law_1.2,0.13432831764221193
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,8,4,power_law_1.2,0.1944537544250488
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,8,4,power_law_1.2,0.27140480041503906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,8,4,power_law_1.01,1.7814515686035157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,8,4,power_law_1.2,0.2755558395385742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,8,4,power_law_1.2,0.2792307281494141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,8,4,power_law_1.2,0.28644351959228515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,8,4,power_law_1.2,0.292410888671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,8,4,power_law_1.01,5.6761755371093745
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,8,4,power_law_1.01,2.5808050537109373
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,8,4,power_law_1.2,0.4412851333618164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,8,4,power_law_1.2,0.4624652862548828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,8,4,power_law_1.2,0.44584190368652343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,8,4,power_law_1.2,0.46448513031005856
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,8,4,power_law_1.2,0.4847615814208984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,8,4,power_law_1.2,0.5245145416259765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,8,4,power_law_1.01,3.408455810546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,8,4,power_law_1.2,0.6381452941894532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,8,4,power_law_1.2,0.7593817901611328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,8,4,power_law_1.2,0.9360588836669923
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,8,4,power_law_1.2,0.4311513519287109
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,8,4,power_law_1.2,0.3081689643859863
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,8,4,power_law_1.2,1.2459417724609376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,8,4,power_law_1.2,0.04883967876434327
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,8,4,power_law_1.2,0.0603110408782959
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,8,4,power_law_1.2,0.08306048393249513
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,8,4,power_law_1.2,0.10513792037963868
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,8,4,power_law_1.2,1.6412185668945312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,8,4,power_law_1.2,0.13127167701721193
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,8,4,power_law_1.2,0.16394367218017578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,8,4,power_law_1.2,0.17189632415771486
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,8,4,power_law_1.2,0.1754342460632324
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,8,4,power_law_1.2,0.18641023635864257
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,8,4,power_law_1.2,0.1981145668029785
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,8,4,power_law_1.2,2.2688832092285156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,8,4,power_law_1.2,0.24497791290283205
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,8,4,power_law_1.2,0.2660979270935059
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,8,4,power_law_1.2,0.24942464828491212
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,8,4,power_law_1.2,0.26390144348144534
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,8,4,power_law_1.2,0.18294912338256836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,8,4,power_law_1.2,3.026025085449219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,8,4,power_law_1.2,0.3007782363891601
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,8,4,power_law_1.2,0.3521331024169922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,8,4,power_law_1.2,0.23744640350341797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,8,4,power_law_1.2,0.39005313873291014
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,8,4,power_law_1.2,0.4923379135131836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,8,4,power_law_1.2,0.5883583831787109
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,8,4,power_law_1.2,0.7923622131347656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,16,2,balanced,0.0509606409072876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,8,4,power_law_1.2,1.012506866455078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,16,2,balanced,0.0736294412612915
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,8,4,power_law_1.2,4.396575927734375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,16,2,balanced,0.2044697570800781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,8,4,power_law_1.2,1.3998118591308593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,16,2,balanced,0.4085286331176758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,8,4,power_law_1.2,1.8455180358886718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,16,2,balanced,0.5305062484741211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,16,2,balanced,0.5424435043334961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,16,2,balanced,0.5471372985839844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,16,2,balanced,0.5603839874267578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,16,2,balanced,0.5749798583984375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,8,4,power_law_1.2,2.6815066528320313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,16,2,balanced,0.5911718368530273
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,16,2,balanced,0.6037977600097657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,16,2,balanced,0.6304064178466797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,16,2,balanced,0.1136575984954834
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,16,2,balanced,0.6982848358154297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,16,2,balanced,0.565844497680664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,16,2,balanced,0.7130290985107421
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,8,4,power_law_1.2,5.933973999023438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,16,2,balanced,0.7376563262939453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,16,2,balanced,0.7837197113037109
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,16,2,balanced,0.8284825897216797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,16,2,balanced,1.273617935180664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,16,2,balanced,1.3722508239746094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,16,2,balanced,0.30888320922851564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,16,2,balanced,1.9865536499023437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,16,2,balanced,2.178040313720703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,16,2,balanced,0.5316198348999024
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,16,2,balanced,3.385511779785156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,16,2,balanced,1.9065023803710937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,16,2,balanced,4.1924505615234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,8,4,power_law_1.2,3.520755310058594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,16,2,balanced,3.559053955078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,16,2,balanced,6.140824584960937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,16,2,balanced,5.196718139648437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,16,2,balanced,7.70213134765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,16,2,balanced,5.13252197265625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,16,2,balanced,5.1862860107421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,16,2,balanced,5.095440673828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,16,2,balanced,5.163463745117188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,16,2,balanced,5.189727783203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,16,2,balanced,5.20248046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,16,2,balanced,4.356716918945312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,16,2,balanced,5.173579711914063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,16,2,balanced,5.209556274414062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,16,2,balanced,4.366632995605469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,16,2,balanced,4.386673889160156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,16,2,balanced,4.4804251098632815
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,16,2,balanced,0.9777356719970702
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,16,2,balanced,4.5458917236328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,16,2,balanced,4.654531860351563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,16,2,balanced,4.771326599121094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,16,2,power_law_1.01,0.07141632080078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,16,2,power_law_1.01,0.10848896026611328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,16,2,power_law_1.01,0.13501824378967284
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,16,2,power_law_1.01,0.20559360504150392
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,16,2,power_law_1.01,0.2753164863586426
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,16,2,power_law_1.01,0.4406284713745118
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,16,2,power_law_1.01,0.46197887420654293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,16,2,power_law_1.01,0.48679424285888667
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,16,2,power_law_1.01,0.490830078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,16,2,power_law_1.01,0.5148876953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,16,2,power_law_1.01,0.5496294403076172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,16,2,power_law_1.01,0.5873088073730469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,16,2,balanced,9.180618286132812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,16,2,balanced,8.922327270507813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,16,2,power_law_1.01,0.6277964782714844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,16,2,power_law_1.01,0.7371148681640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,16,2,power_law_1.01,0.6733158111572266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,16,2,power_law_1.01,0.6997644805908203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,16,2,power_law_1.01,0.743157730102539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,16,2,power_law_1.01,0.8465164947509767
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,16,2,power_law_1.01,0.9780198669433593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,16,2,power_law_1.01,1.17212158203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,16,2,power_law_1.01,1.4254106140136718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,16,2,power_law_1.01,1.842027587890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,16,2,power_law_1.01,2.332769317626953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,16,2,balanced,13.54299560546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,16,2,power_law_1.01,0.514672622680664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,16,2,balanced,14.274885253906252
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,16,2,power_law_1.01,3.2519256591796877
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,16,2,power_law_1.01,0.9583526611328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,16,2,power_law_1.01,1.294958038330078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,16,2,power_law_1.01,1.9015245056152346
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,16,2,power_law_1.01,4.249338989257812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,16,2,power_law_1.01,2.664989318847656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,16,2,power_law_1.01,6.226187744140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,16,2,power_law_1.01,4.155603332519531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,16,2,power_law_1.01,4.30935302734375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,16,2,power_law_1.01,8.104050903320312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,16,2,power_law_1.01,4.244547729492187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,16,2,power_law_1.01,4.5239999389648435
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,16,2,power_law_1.01,4.473641052246093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,16,2,balanced,22.852910156249997
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,16,2,power_law_1.01,4.548116455078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,16,2,power_law_1.01,4.712732238769531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,16,2,power_law_1.01,4.598984985351562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,16,2,power_law_1.01,4.80429443359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,16,2,power_law_1.01,4.021191711425781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,16,2,power_law_1.01,4.053466796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,16,2,power_law_1.01,4.138009643554687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,16,2,power_law_1.01,4.549726867675782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,16,2,balanced,28.29180908203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,16,2,power_law_1.01,4.8551092529296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,16,2,power_law_1.01,5.699759521484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,16,2,power_law_1.2,0.07119743824005127
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,16,2,power_law_1.2,0.09376640319824218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,16,2,power_law_1.2,0.13502464294433594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,16,2,power_law_1.01,6.595132446289062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,16,2,power_law_1.2,0.18633472442626953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,16,2,power_law_1.2,0.2814425659179688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,16,2,power_law_1.2,0.4475737762451172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,16,2,power_law_1.2,0.46118015289306646
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,16,2,power_law_1.2,0.47569152832031253
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,16,2,power_law_1.01,8.102769775390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,16,2,power_law_1.2,0.49647743225097657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,16,2,power_law_1.2,0.5174771118164062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,16,2,power_law_1.2,0.5684851074218751
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,16,2,power_law_1.2,0.5971136093139648
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,16,2,power_law_1.2,0.6688934326171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,16,2,power_law_1.2,0.7501606750488281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,16,2,power_law_1.2,0.6958604431152343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,16,2,power_law_1.01,9.757868652343749
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,16,2,power_law_1.2,0.7138214111328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,16,2,power_law_1.2,0.7615615844726562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,16,2,power_law_1.2,0.8527603149414062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,16,2,power_law_1.2,0.9866252899169922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,16,2,power_law_1.2,1.1941824340820313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,16,2,power_law_1.2,1.4620979309082032
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,16,2,power_law_1.2,1.8677325439453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,16,2,power_law_1.01,12.888414306640623
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,16,2,power_law_1.2,2.397227478027344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,16,2,power_law_1.2,0.5199539184570312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,16,2,power_law_1.2,3.3336587524414063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,16,2,power_law_1.2,0.8237120056152344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,16,2,power_law_1.2,1.2344012451171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,16,2,power_law_1.2,4.3027340698242185
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,16,2,power_law_1.2,1.9110873413085936
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,16,2,power_law_1.01,16.13984375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,16,2,power_law_1.2,2.6108084106445313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,16,2,power_law_1.2,6.394379272460937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,16,2,power_law_1.2,4.1450958251953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,16,2,power_law_1.2,4.276719360351562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,16,2,power_law_1.2,4.439875793457031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,16,2,power_law_1.2,4.4443994140625005
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,16,2,power_law_1.2,4.419774780273437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,16,2,power_law_1.2,4.680405883789062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,16,2,power_law_1.01,23.48973388671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,16,2,power_law_1.2,4.766147766113281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,16,2,power_law_1.2,4.866141357421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,16,2,power_law_1.2,4.1711578369140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,16,2,power_law_1.2,5.079246215820312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,16,2,power_law_1.2,8.076437988281251
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,16,2,power_law_1.2,4.222616271972656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,16,2,power_law_1.2,4.446749572753906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,16,2,power_law_1.01,29.100429687500004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,16,2,power_law_1.2,5.136323852539062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,16,2,power_law_1.2,4.865761413574218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,16,2,power_law_1.2,5.883944702148438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,16,2,power_law_1.2,6.8734326171874995
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,32,1,balanced,0.0776806402206421
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,32,1,balanced,0.16274496078491213
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,16,2,power_law_1.2,8.411618652343751
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,32,1,balanced,0.27873151779174804
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,16,2,power_law_1.2,10.3461767578125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,32,1,balanced,0.10673088073730468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,32,1,balanced,0.5526700973510742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,32,1,balanced,0.7499539184570312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,32,1,balanced,0.7818316650390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,32,1,balanced,0.7607705688476563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,32,1,balanced,0.7603596496582031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,32,1,balanced,0.7732498931884766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,32,1,balanced,0.7888864135742187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,32,1,balanced,0.7999545288085937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,16,2,power_law_1.2,13.49225830078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,32,1,balanced,0.8117164611816406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,32,1,balanced,0.8317203521728516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,32,1,balanced,0.8294502258300781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,32,1,balanced,1.1078482818603514
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,16,2,power_law_1.2,16.917354736328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,32,1,balanced,0.8362252807617188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,32,1,balanced,0.8692415618896485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,32,1,balanced,1.4325779724121095
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,32,1,balanced,1.5068620300292968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,32,1,balanced,1.5708934020996093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,32,1,balanced,2.445378570556641
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,32,1,balanced,3.8436761474609376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,32,1,balanced,2.2824473571777344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,16,2,power_law_1.2,24.190712890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,32,1,balanced,0.3465651321411133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,32,1,balanced,4.755210876464844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,32,1,balanced,7.012525024414063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,32,1,balanced,0.6147347259521484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,32,1,balanced,1.1206905364990234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,32,1,balanced,8.7917578125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,32,1,balanced,2.180431365966797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,16,2,power_law_1.2,30.2066357421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,32,1,balanced,6.151973266601563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,32,1,balanced,5.8558203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,32,1,balanced,6.05999755859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,32,1,balanced,6.080208129882813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,32,1,balanced,6.070836181640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,32,1,balanced,6.154801025390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,32,1,balanced,5.90778564453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,32,1,balanced,6.166452026367187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,32,1,balanced,6.106627197265626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,32,1,balanced,6.130792236328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,32,1,balanced,4.238568420410156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,32,1,balanced,6.135444946289063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,32,1,balanced,6.15798095703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,32,1,balanced,6.080110473632812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,32,1,balanced,6.05898193359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,32,1,balanced,6.314752807617188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,32,1,balanced,6.3734936523437495
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,32,1,power_law_1.01,0.07770368099212646
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,32,1,power_law_1.01,0.10666432380676269
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,32,1,power_law_1.01,0.16209728240966798
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,32,1,power_law_1.01,0.25492736816406253
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,32,1,power_law_1.01,0.5973977661132812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,32,1,power_law_1.01,0.3614604949951172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,32,1,power_law_1.01,0.6332633590698242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,32,1,power_law_1.01,0.6546399688720703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,32,1,power_law_1.01,0.6754681396484374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,32,1,power_law_1.01,0.6956473541259766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,32,1,power_law_1.01,0.7429183959960938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,32,1,balanced,12.028211669921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,32,1,power_law_1.01,0.8018450927734374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,32,1,balanced,12.56992919921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,32,1,power_law_1.01,0.8638591766357422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,32,1,power_law_1.01,1.0661491394042968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,32,1,power_law_1.01,1.1881696319580077
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,32,1,power_law_1.01,1.401901397705078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,32,1,power_law_1.01,0.9750739288330077
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,32,1,power_law_1.01,0.9655276489257811
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,32,1,power_law_1.01,1.1053202819824217
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,32,1,power_law_1.01,1.345638427734375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,32,1,power_law_1.01,2.1134439086914063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,32,1,power_law_1.01,2.6154330444335936
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,32,1,power_law_1.01,1.603822021484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,32,1,balanced,18.19548828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,32,1,power_law_1.01,0.3453964614868164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,32,1,power_law_1.01,3.6384811401367188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,32,1,balanced,19.058912353515627
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,32,1,power_law_1.01,0.6149856185913085
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,32,1,power_law_1.01,1.1093145751953126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,32,1,power_law_1.01,4.700488891601562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,32,1,power_law_1.01,6.857968139648437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,32,1,power_law_1.01,4.469783630371094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,32,1,power_law_1.01,4.803795166015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,32,1,power_law_1.01,9.033052978515624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,32,1,power_law_1.01,2.7287066650390623
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,32,1,power_law_1.01,5.105355529785156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,32,1,power_law_1.01,5.198499755859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,32,1,power_law_1.01,5.298380737304687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,32,1,balanced,31.035222167968747
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,32,1,power_law_1.01,5.066010742187499
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,32,1,power_law_1.01,5.368353271484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,32,1,power_law_1.01,5.4086572265625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,32,1,power_law_1.01,1.940498504638672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,32,1,power_law_1.01,5.6142431640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,32,1,power_law_1.01,5.774144897460937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,32,1,power_law_1.01,6.1146215820312495
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,32,1,power_law_1.01,6.578361206054687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,32,1,power_law_1.01,6.064707641601562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,32,1,balanced,37.7265380859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,32,1,power_law_1.01,6.498037719726563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,32,1,power_law_1.01,7.503107299804688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,32,1,power_law_1.2,0.07746175765991212
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,32,1,power_law_1.2,0.10689920425415038
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,32,1,power_law_1.2,0.16215871810913085
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,32,1,power_law_1.01,8.660322265625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,32,1,power_law_1.2,0.33263679504394533
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,32,1,power_law_1.2,0.5939507293701172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,32,1,power_law_1.01,10.75413818359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,32,1,power_law_1.2,0.6325555038452149
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,32,1,power_law_1.2,0.23856832504272463
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,32,1,power_law_1.2,0.6492908477783204
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,32,1,power_law_1.2,0.6819737243652344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,32,1,power_law_1.2,0.7128217315673828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,32,1,power_law_1.2,0.758560028076172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,32,1,power_law_1.2,0.8394547271728516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,32,1,power_law_1.2,0.8947756958007812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,32,1,power_law_1.2,1.0059865570068358
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,32,1,power_law_1.2,1.1040115356445312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,32,1,power_law_1.01,12.853791503906251
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,32,1,power_law_1.2,1.21238525390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,32,1,power_law_1.2,1.0020076751708984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,32,1,power_law_1.2,1.4388064575195312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,32,1,power_law_1.2,1.3900915527343751
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,32,1,power_law_1.2,1.6458969116210938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,32,1,power_law_1.2,1.1431846618652344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,32,1,power_law_1.2,2.1509011840820316
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,32,1,power_law_1.2,2.6624871826171876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,32,1,power_law_1.01,17.234066162109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,32,1,power_law_1.2,0.34292606353759764
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,32,1,power_law_1.2,0.6210764694213867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,32,1,power_law_1.2,3.708619384765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,32,1,power_law_1.2,1.1272729492187499
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,32,1,power_law_1.2,4.7428890991210935
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,32,1,power_law_1.01,21.26021484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,32,1,power_law_1.2,2.464789733886719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,32,1,power_law_1.2,4.444302062988281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,32,1,power_law_1.2,4.758040466308594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,32,1,power_law_1.2,5.0000109863281255
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,32,1,power_law_1.2,6.910105590820312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,32,1,power_law_1.01,29.736064453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,32,1,power_law_1.2,5.122503662109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,32,1,power_law_1.2,5.233088989257813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,32,1,power_law_1.2,5.478976440429688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,32,1,power_law_1.2,5.503137817382813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,32,1,power_law_1.2,5.386249389648437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,32,1,power_law_1.2,1.7882284545898437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,32,1,power_law_1.2,5.947013549804687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,32,1,power_law_1.2,6.138549194335938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,32,1,power_law_1.01,38.39399658203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,32,1,power_law_1.2,6.500991210937499
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,32,1,power_law_1.2,6.8569873046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,32,1,power_law_1.2,9.082421264648437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,32,1,power_law_1.2,6.428508911132812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,32,1,power_law_1.2,6.991181030273438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,32,1,power_law_1.2,8.001317749023437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,8,8,balanced,0.042154240608215335
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,8,8,balanced,0.04252927780151367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,8,8,balanced,0.04221183776855469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,8,8,balanced,0.07503488063812255
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,8,8,balanced,0.07584127902984619
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,8,8,balanced,0.07577087879180908
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,8,8,balanced,0.07726336002349854
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,8,8,balanced,0.07564671993255614
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,8,8,balanced,0.07764863967895508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,8,8,balanced,0.0781388807296753
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,8,8,balanced,0.07829631805419922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,8,8,balanced,0.07871615886688232
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,32,1,power_law_1.2,8.917879638671874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,8,8,balanced,0.07902592182159424
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,8,8,balanced,0.08263680458068848
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,8,8,balanced,0.08250240325927735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,8,8,balanced,0.08437631607055665
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,8,8,balanced,0.08658047676086425
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,8,8,balanced,0.09322239875793456
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,8,8,balanced,0.09756671905517578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,8,8,balanced,0.10565759658813476
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,8,8,balanced,0.1032921600341797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,8,8,balanced,0.18045183181762697
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,8,8,balanced,0.2084275245666504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,8,8,balanced,0.3084275245666504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,8,8,balanced,0.37490047454833986
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,8,8,balanced,0.5431232070922851
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,8,8,balanced,0.026828799247741696
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,8,8,balanced,0.7123788452148437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,8,8,balanced,0.026878719329833982
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,8,8,balanced,0.026657280921936033
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,8,8,balanced,0.04618879795074463
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,8,8,balanced,0.046419200897216795
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,8,8,balanced,0.04641151905059814
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,8,8,balanced,0.04624000072479248
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,8,8,balanced,0.04645503997802734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,8,8,balanced,0.046855678558349614
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,8,8,balanced,0.04698880195617676
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,8,8,balanced,0.04712831974029541
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,8,8,balanced,0.04755712032318115
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,8,8,balanced,0.04772863864898681
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,8,8,balanced,0.04841216087341309
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,32,1,power_law_1.2,10.946641845703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,8,8,balanced,0.048902401924133296
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,8,8,balanced,0.04938111782073974
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,8,8,balanced,0.05076608180999755
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,8,8,balanced,0.05299456119537353
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,8,8,balanced,0.05549824237823486
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,8,8,balanced,0.06599040031433105
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,8,8,balanced,0.06072192192077637
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,8,8,balanced,0.07073023796081543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,8,8,balanced,0.08380415916442871
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,8,8,balanced,0.12233728408813475
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,8,8,balanced,0.14447615623474122
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,8,8,balanced,0.20866304397583008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,8,8,power_law_1.01,0.04106624126434326
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,8,8,balanced,0.2705971145629883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,8,8,power_law_1.01,0.07074944019317628
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,8,8,power_law_1.01,0.07261439800262451
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,8,8,power_law_1.01,0.07725056171417236
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,8,8,power_law_1.01,0.07615488052368165
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,8,8,power_law_1.01,0.07504767894744872
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,8,8,power_law_1.01,0.0806719970703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,8,8,power_law_1.01,0.07889408111572266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,8,8,power_law_1.01,0.08349568367004394
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,8,8,power_law_1.01,0.08234623908996583
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,8,8,power_law_1.01,0.08766976356506348
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,8,8,power_law_1.01,0.08960767745971679
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,8,8,power_law_1.01,0.09113087654113769
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,8,8,power_law_1.01,0.09666175842285156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,8,8,power_law_1.01,0.0969382381439209
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,8,8,power_law_1.01,0.10211071968078614
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,8,8,power_law_1.01,0.10774784088134766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,32,1,power_law_1.2,13.285155029296874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,8,8,power_law_1.01,0.16925695419311523
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,8,8,power_law_1.01,0.20802688598632812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,8,8,power_law_1.01,0.3631999969482422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,8,8,power_law_1.01,0.288852481842041
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,8,8,power_law_1.01,0.5019583892822266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,8,8,power_law_1.01,0.6911334228515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,32,1,power_law_1.2,17.45389892578125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,8,8,power_law_1.01,0.027178239822387696
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,8,8,power_law_1.01,0.044618239402771
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,8,8,power_law_1.01,0.04583424091339111
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,8,8,power_law_1.01,0.045715198516845704
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,8,8,power_law_1.01,0.04608767986297607
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,8,8,power_law_1.01,0.04647295951843262
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,8,8,power_law_1.01,0.04681727886199951
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,8,8,power_law_1.01,0.04719871997833252
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,8,8,power_law_1.01,0.04798463821411133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,8,8,power_law_1.01,0.04816768169403076
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,8,8,power_law_1.01,0.04952703952789307
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,8,8,power_law_1.01,0.05020544052124023
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,8,8,power_law_1.01,0.051614718437194826
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,8,8,power_law_1.01,1.0166937255859376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,8,8,power_law_1.01,0.05313536167144776
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,8,8,power_law_1.01,0.0580838394165039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,8,8,power_law_1.01,0.059413762092590326
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,8,8,power_law_1.01,0.06284031867980958
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,8,8,power_law_1.01,0.06740608215332031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,8,8,power_law_1.01,0.0787340784072876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,8,8,power_law_1.01,0.1133619213104248
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,8,8,power_law_1.01,0.1345792007446289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,8,8,power_law_1.01,0.1899660873413086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,32,1,power_law_1.2,21.60732666015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,8,8,power_law_1.01,0.25113088607788087
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,8,8,power_law_1.01,1.922144012451172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,8,8,power_law_1.01,1.3950157165527344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,8,8,power_law_1.2,0.04154880046844482
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,8,8,power_law_1.2,0.0751308822631836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,8,8,power_law_1.2,0.07394559860229492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,8,8,power_law_1.01,0.47917823791503905
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,8,8,power_law_1.2,0.07107967853546142
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,8,8,power_law_1.2,0.07270400047302246
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,8,8,power_law_1.01,2.5533311462402346
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,8,8,power_law_1.2,0.07464960098266601
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,8,8,power_law_1.2,0.07986944198608399
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,8,8,power_law_1.2,0.0803980827331543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,8,8,power_law_1.2,0.08233087539672852
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,8,8,power_law_1.2,0.08742527961730957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,8,8,power_law_1.2,0.08872063636779785
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,8,8,power_law_1.2,0.09352959632873535
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,8,8,power_law_1.2,0.09613183975219727
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,8,8,power_law_1.2,0.09941887855529785
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,8,8,power_law_1.2,0.08463359832763671
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,8,8,power_law_1.2,0.10476160049438477
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,8,8,power_law_1.2,0.10549759864807129
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,8,8,power_law_1.01,0.32756225585937504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,8,8,power_law_1.2,0.1622118377685547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,8,8,power_law_1.2,0.29315200805664066
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,8,8,power_law_1.2,0.2062604713439941
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,8,8,power_law_1.2,0.3399859237670898
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,8,8,power_law_1.2,0.5279756927490235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,8,8,power_law_1.01,0.6636326599121094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,8,8,power_law_1.01,0.9054271697998046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,8,8,power_law_1.2,0.028663039207458496
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,8,8,power_law_1.2,0.04654208183288574
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,8,8,power_law_1.2,0.04658688068389892
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,8,8,power_law_1.2,0.04662911891937256
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,8,8,power_law_1.2,0.047029762268066405
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,8,8,power_law_1.2,0.04752255916595459
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,8,8,power_law_1.2,0.04711552143096924
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,8,8,power_law_1.2,0.04775040149688721
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,8,8,power_law_1.2,0.048162560462951656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,8,8,power_law_1.2,0.049030399322509764
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,8,8,power_law_1.2,0.04958720207214355
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,8,8,power_law_1.2,0.6993651580810547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,8,8,power_law_1.2,0.051294717788696285
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,32,1,power_law_1.2,30.016767578125002
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,8,8,power_law_1.2,0.05286655902862549
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,8,8,power_law_1.2,0.055299839973449706
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,8,8,power_law_1.2,0.06293248176574708
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,8,8,power_law_1.2,0.06268544197082519
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,8,8,power_law_1.2,0.06316800117492676
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,8,8,power_law_1.2,0.0680140781402588
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,8,8,power_law_1.2,0.08047360420227051
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,8,8,power_law_1.2,0.11498111724853514
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,8,8,power_law_1.2,0.1334233570098877
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,8,8,power_law_1.2,0.19422336578369143
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,8,8,power_law_1.2,2.0108493041992186
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,8,8,power_law_1.2,1.046264343261719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,8,8,power_law_1.2,0.2542323112487793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,16,4,balanced,0.023809280395507813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,16,4,balanced,0.023956480026245116
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,16,4,balanced,0.037080318927764894
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,16,4,balanced,0.06248832225799561
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,16,4,balanced,0.06334335803985595
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,16,4,balanced,0.06612095832824708
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,16,4,balanced,0.0657907199859619
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,16,4,balanced,0.06759552001953126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,16,4,balanced,0.0672870397567749
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,16,4,balanced,0.06891136169433594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,16,4,balanced,0.06987520217895507
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,16,4,balanced,0.07186431884765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,16,4,balanced,0.07365503787994385
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,16,4,balanced,0.07553408145904542
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,16,4,balanced,0.07873663902282715
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,16,4,balanced,0.08112000465393067
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,16,4,balanced,0.08490240097045898
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,16,4,balanced,0.09455231666564942
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,16,4,balanced,0.08819583892822266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,16,4,balanced,0.10469887733459474
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,16,4,balanced,0.117706241607666
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,16,4,balanced,0.17479040145874022
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,16,4,balanced,0.22086271286010742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,16,4,balanced,0.3259238433837891
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,16,4,balanced,0.4046783828735352
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,8,8,power_law_1.2,0.3424652862548828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,16,4,balanced,0.5868121719360351
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,16,4,balanced,0.02450687885284424
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,16,4,balanced,0.7710809326171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,16,4,balanced,0.024453120231628416
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,16,4,balanced,0.026406400203704834
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,16,4,balanced,0.0413375997543335
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,16,4,balanced,0.04197887897491455
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,16,4,balanced,0.041740798950195314
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,16,4,balanced,0.042455039024353027
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,16,4,balanced,0.04209792137145996
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,16,4,balanced,0.042813439369201664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,16,4,balanced,0.04244736194610595
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,16,4,balanced,0.043415040969848634
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,16,4,balanced,0.043393278121948244
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,16,4,balanced,0.04443647861480713
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,16,4,balanced,0.045240321159362794
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,16,4,balanced,0.0473356819152832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,16,4,balanced,0.048846077919006345
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,16,4,balanced,0.05105023860931397
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,16,4,balanced,0.06076543807983399
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,16,4,balanced,0.055092477798461915
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,16,4,balanced,0.06163072109222412
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,16,4,balanced,0.07242879867553711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,16,4,balanced,0.08596351623535156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,16,4,balanced,0.10542976379394531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,16,4,balanced,0.1576358413696289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,16,4,balanced,0.18762752532958984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,8,8,power_law_1.2,1.3921189880371094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,16,4,balanced,0.27313024520874024
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,16,4,power_law_1.01,0.029250559806823728
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,16,4,power_law_1.01,0.036989440917968755
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,16,4,balanced,0.36151168823242186
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,16,4,power_law_1.01,0.056541438102722164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,8,8,power_law_1.2,0.4856294250488281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,16,4,power_law_1.01,0.0633958387374878
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,16,4,power_law_1.01,0.06229631900787354
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,16,4,power_law_1.01,0.06932608127593995
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,16,4,power_law_1.01,0.0670143985748291
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,16,4,power_law_1.01,0.07355135917663574
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,16,4,power_law_1.01,0.07397759914398193
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,16,4,power_law_1.01,0.07570816040039062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,16,4,power_law_1.01,0.07926527976989746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,16,4,power_law_1.01,0.08083840370178222
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,16,4,power_law_1.01,0.08752384185791015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,16,4,power_law_1.01,0.08871808052062988
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,8,8,power_law_1.2,2.6758630371093752
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,16,4,power_law_1.01,0.09401215553283691
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,16,4,power_law_1.01,0.0985971164703369
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,16,4,power_law_1.01,0.06665599822998047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,16,4,power_law_1.01,0.11152640342712403
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,16,4,power_law_1.01,0.1492121601104736
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,16,4,power_law_1.01,0.2295987129211426
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,16,4,power_law_1.01,0.205611515045166
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,16,4,power_law_1.01,0.3719411087036133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,16,4,power_law_1.01,0.45798526763916014
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,16,4,power_law_1.01,0.024435200691223145
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,16,4,power_law_1.01,0.02583679914474487
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,16,4,power_law_1.01,0.041543679237365724
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,16,4,power_law_1.01,0.04160511970520019
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,16,4,power_law_1.01,0.041831679344177246
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,16,4,power_law_1.01,0.042214398384094236
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,16,4,power_law_1.01,0.0425651216506958
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,16,4,power_law_1.01,0.04309375762939453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,16,4,power_law_1.01,0.04349440097808838
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,16,4,power_law_1.01,0.04403071880340576
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,16,4,power_law_1.01,0.044916481971740724
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,32,1,power_law_1.2,39.2012158203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,8,8,power_law_1.2,0.6898342132568359
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,16,4,power_law_1.01,0.048078079223632816
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,16,4,power_law_1.01,0.04849535942077637
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,16,4,power_law_1.01,0.055567359924316405
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,16,4,power_law_1.01,0.059293441772460945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,16,4,power_law_1.01,0.05781248092651368
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,16,4,power_law_1.01,0.06253183841705322
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,16,4,power_law_1.01,0.04660223960876465
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,8,8,power_law_1.2,0.9564620971679687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,16,4,power_law_1.01,0.06835328102111816
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,16,4,power_law_1.01,0.7582514953613282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,16,4,power_law_1.01,0.0953446388244629
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,16,4,power_law_1.01,0.10945792198181152
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,16,4,power_law_1.01,0.1486950397491455
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,16,4,power_law_1.01,0.20404991149902343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,16,4,power_law_1.2,0.03240832090377808
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,16,4,power_law_1.2,0.029762558937072754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,16,4,power_law_1.2,0.04882815837860107
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,16,4,power_law_1.2,0.06366591930389405
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,16,4,power_law_1.2,0.06251264095306397
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,16,4,power_law_1.2,0.06714752197265625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,16,4,power_law_1.01,0.3596364974975586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,16,4,power_law_1.2,0.06734848022460938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,16,4,power_law_1.2,0.06748799800872803
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,16,4,power_law_1.2,0.07055359840393066
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,16,4,power_law_1.2,0.07273983955383301
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,16,4,power_law_1.2,0.0753868818283081
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,16,4,power_law_1.01,1.6553575134277345
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,16,4,power_law_1.2,0.07860735893249512
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,16,4,power_law_1.2,0.08063360214233398
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,16,4,power_law_1.2,0.0880128002166748
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,16,4,power_law_1.2,0.09006848335266113
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,16,4,power_law_1.2,0.0936563205718994
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,16,4,power_law_1.01,0.965582046508789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,16,4,power_law_1.2,0.09811967849731446
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,16,4,power_law_1.2,0.11111424446105958
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,16,4,power_law_1.01,0.31664384841918947
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,16,4,power_law_1.2,0.20278783798217775
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,16,4,power_law_1.2,0.14700160026550294
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,16,4,power_law_1.2,0.23456512451171876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,16,4,power_law_1.01,0.4920857620239258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,16,4,power_law_1.01,1.28606201171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,16,4,power_law_1.2,0.024807679653167724
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,16,4,power_law_1.2,0.02610687971115112
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,16,4,power_law_1.2,0.041843199729919435
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,16,4,power_law_1.2,0.35213695526123046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,16,4,power_law_1.2,0.04187903881072998
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,16,4,power_law_1.2,0.041969919204711915
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,16,4,power_law_1.2,0.04254464149475098
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,16,4,power_law_1.2,0.042516479492187495
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,16,4,power_law_1.2,0.043357439041137695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,16,4,power_law_1.2,0.0435481595993042
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,16,4,power_law_1.2,0.04428287982940674
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,16,4,power_law_1.2,0.04497791767120361
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,16,4,power_law_1.2,0.04605696201324463
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,16,4,power_law_1.2,0.04673151969909668
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,16,4,power_law_1.2,0.04900735855102539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,16,4,power_law_1.2,0.05666048049926757
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,16,4,power_law_1.2,0.05980927944183349
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,16,4,power_law_1.2,0.05818880081176758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,16,4,power_law_1.01,0.7209945678710937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,16,4,power_law_1.2,0.4734323120117187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,16,4,power_law_1.2,0.0627123212814331
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,16,4,power_law_1.2,0.07087744235992431
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,16,4,power_law_1.2,0.09659135818481446
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,16,4,power_law_1.2,0.1091648006439209
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,16,4,power_law_1.2,0.14815872192382812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,16,4,power_law_1.2,0.20490495681762697
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,16,4,power_law_1.2,0.7666483306884766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,32,2,balanced,0.022239999771118162
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,32,2,balanced,0.024075520038604737
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,32,2,balanced,0.03734272003173828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,32,2,balanced,0.06268159866333008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,32,2,balanced,0.06295551776885985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,32,2,balanced,0.06379007816314697
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,16,4,power_law_1.2,0.3543743896484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,32,2,balanced,0.06445824146270752
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,16,4,power_law_1.2,0.30869760513305666
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,32,2,balanced,0.06518400192260743
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,32,2,balanced,0.06531328201293946
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,32,2,balanced,0.0681331205368042
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,32,2,balanced,0.07004799842834472
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,32,2,balanced,0.0721343994140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,32,2,balanced,0.07461631774902344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,32,2,balanced,0.08017024040222168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,32,2,balanced,0.08327551841735839
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,32,2,balanced,0.08709759712219238
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,32,2,balanced,0.08046719551086426
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,32,2,balanced,0.09252479553222656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,32,2,balanced,0.09980928421020507
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,32,2,balanced,0.1179353618621826
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,32,2,balanced,0.13386624336242675
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,32,2,balanced,0.20013055801391602
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,16,4,power_law_1.2,1.6652365112304686
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,32,2,balanced,0.25453439712524417
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,32,2,balanced,0.37552513122558595
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,32,2,balanced,0.47159423828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,32,2,balanced,0.023427839279174807
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,32,2,balanced,0.025105919837951664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,32,2,balanced,0.69193603515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,32,2,balanced,0.0403110408782959
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,32,2,balanced,0.913144302368164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,32,2,balanced,0.04077824115753174
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,32,2,balanced,0.04070144176483154
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,32,2,balanced,0.041509118080139164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,32,2,balanced,0.04129024028778076
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,32,2,balanced,0.042175998687744146
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,32,2,balanced,0.042707200050354006
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,32,2,balanced,0.04454016208648682
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,32,2,balanced,0.04552576065063477
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,32,2,balanced,0.05131648063659668
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,32,2,balanced,0.05466752052307129
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,32,2,balanced,0.039997439384460445
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,32,2,balanced,0.05708032131195069
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,32,2,balanced,0.05274367809295655
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,32,2,balanced,0.0628223991394043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,32,2,balanced,0.06871679782867432
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,32,2,balanced,0.07936255931854248
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,32,2,balanced,0.09519871711730957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,32,2,balanced,0.12760576248168945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,32,2,balanced,0.1535871982574463
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,32,2,balanced,0.23255168914794924
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,16,4,power_law_1.2,0.5027891159057617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,32,2,balanced,0.291333122253418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,32,2,power_law_1.01,0.0278003191947937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,32,2,power_law_1.01,0.03487103939056396
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,32,2,balanced,0.425794563293457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,32,2,power_law_1.01,0.03164671897888184
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,32,2,power_law_1.01,0.06098688125610352
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,32,2,power_law_1.01,0.06313344001770019
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,32,2,balanced,0.5615372848510742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,32,2,power_law_1.01,0.06235904216766357
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,32,2,power_law_1.01,0.06372992038726807
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,32,2,power_law_1.01,0.06633984088897706
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,32,2,power_law_1.01,0.06705408096313477
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,32,2,power_law_1.01,0.06765952110290527
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,32,2,power_law_1.01,0.07057919979095459
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,32,2,power_law_1.01,0.07250688076019288
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,32,2,power_law_1.01,0.07554175853729247
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,32,2,power_law_1.01,0.08288127899169921
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,32,2,power_law_1.01,0.08390144348144532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,32,2,power_law_1.01,0.08675711631774903
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,32,2,power_law_1.01,0.08807552337646485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,32,2,power_law_1.01,0.0987110424041748
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,32,2,balanced,0.027162880897521974
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,16,4,power_law_1.2,1.250360336303711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,32,2,power_law_1.01,0.11203840255737305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,32,2,power_law_1.01,0.15988351821899413
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,32,2,power_law_1.01,0.13311871528625488
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,16,4,power_law_1.2,0.9737177276611328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,16,4,power_law_1.2,0.7138009643554687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,32,2,power_law_1.01,0.025552639961242674
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,32,2,power_law_1.01,0.24526079177856444
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,32,2,power_law_1.01,0.861269760131836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,32,2,power_law_1.01,0.02375296115875244
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,32,2,power_law_1.01,0.3360627365112305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,32,2,power_law_1.01,0.04108672142028809
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,32,2,power_law_1.01,0.026246399879455568
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,32,2,power_law_1.01,0.48636032104492183
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,32,2,power_law_1.01,0.042891521453857426
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,32,2,power_law_1.01,0.03997440099716186
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,32,2,power_law_1.01,0.045192961692810056
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,32,2,power_law_1.01,0.04612095832824707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,32,2,power_law_1.01,0.049698557853698734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,32,2,power_law_1.01,0.0554585599899292
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,32,2,power_law_1.01,0.05671296119689941
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,32,2,power_law_1.01,0.056272640228271484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,32,2,power_law_1.01,0.040087041854858396
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,32,2,power_law_1.01,0.06486144065856933
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,32,2,power_law_1.01,0.06748032093048095
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,32,2,power_law_1.01,0.08156928062438965
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,32,2,power_law_1.01,0.0414463996887207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,32,2,power_law_1.01,0.04206975936889649
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,32,2,power_law_1.01,0.15714303970336913
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,32,2,power_law_1.01,0.10033280372619628
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,32,2,power_law_1.01,0.6173747253417969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,32,2,power_law_1.01,0.043610877990722656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,32,2,power_law_1.2,0.027336959838867185
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,32,2,power_law_1.2,0.03481343984603882
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,32,2,power_law_1.2,0.030999040603637694
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,32,2,power_law_1.2,0.058246397972106935
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,32,2,power_law_1.2,0.05956352233886719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,32,2,power_law_1.2,0.061478400230407716
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,32,2,power_law_1.2,0.06345088005065917
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,32,2,power_law_1.01,0.04068480014801025
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,32,2,power_law_1.2,0.06600319862365722
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,32,2,power_law_1.2,0.06649727821350097
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,32,2,power_law_1.2,0.06817152023315429
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,32,2,power_law_1.2,0.06876031875610351
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,32,2,power_law_1.2,0.07302783966064454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,32,2,power_law_1.01,1.0810931396484373
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,32,2,power_law_1.2,0.07655295848846436
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,32,2,power_law_1.2,0.08562047958374024
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,32,2,power_law_1.2,0.0815014362335205
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,32,2,power_law_1.01,0.1885094451904297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,32,2,power_law_1.2,0.0867404842376709
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,32,2,power_law_1.2,0.08815103530883789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,32,2,power_law_1.2,0.10800767898559571
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,32,2,power_law_1.2,0.09856639862060547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,32,2,power_law_1.01,0.34988033294677734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,32,2,power_law_1.2,0.13229567527770997
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,32,2,power_law_1.2,0.16308992385864257
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,32,2,power_law_1.01,0.2711680030822754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,32,2,power_law_1.01,0.6426112365722656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,32,2,power_law_1.01,0.48853759765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,32,2,power_law_1.2,0.023682560920715332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,32,2,power_law_1.2,0.2593600082397461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,32,2,power_law_1.2,0.03997567892074585
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,32,2,power_law_1.2,0.04007423877716064
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,32,2,power_law_1.2,0.04041855812072754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,32,2,power_law_1.2,0.0410368013381958
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,32,2,power_law_1.2,0.04143743991851807
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,32,2,power_law_1.2,0.04198016166687012
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,32,2,power_law_1.2,0.04250495910644531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,32,2,power_law_1.2,0.043889918327331544
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,32,2,power_law_1.2,0.04462463855743408
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,32,2,power_law_1.2,0.046743040084838865
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,32,2,power_law_1.2,0.05164671897888183
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,32,2,power_law_1.2,0.055287041664123536
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,32,2,power_law_1.2,0.05687679767608642
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,32,2,power_law_1.2,0.025195519924163818
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,32,2,power_law_1.2,0.026467840671539306
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,32,2,power_law_1.2,0.056940798759460454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,32,2,power_law_1.2,0.3251174545288086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,32,2,power_law_1.2,0.06443264007568358
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,32,2,power_law_1.2,0.0682035207748413
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,32,2,power_law_1.2,0.0821132755279541
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,32,2,power_law_1.2,0.15680768013000487
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,32,2,power_law_1.2,0.09743488311767579
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,32,2,power_law_1.2,0.49488254547119137
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,32,2,power_law_1.2,0.8894400024414063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,8,8,balanced,0.061098241806030275
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,8,8,balanced,0.061053438186645506
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,32,2,power_law_1.2,0.18799232482910155
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,8,8,balanced,0.06098944187164307
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,8,8,balanced,0.1116096019744873
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,8,8,balanced,0.11237631797790529
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,8,8,balanced,0.11194368362426758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,8,8,balanced,0.11245439529418946
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,8,8,balanced,0.11354368209838868
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,8,8,balanced,0.11332480430603027
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,8,8,balanced,0.11236096382141114
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,8,8,balanced,0.11595904350280761
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,8,8,balanced,0.11673472404479981
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,8,8,balanced,0.1186521625518799
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,8,8,balanced,0.11823231697082519
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,8,8,balanced,0.11884927749633789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,8,8,balanced,0.11403391838073731
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,8,8,balanced,0.12508928298950195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,8,8,balanced,0.12882304191589355
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,8,8,balanced,0.1312384033203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,8,8,balanced,0.1581375980377197
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,8,8,balanced,0.16550655364990235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,8,8,balanced,0.27404415130615234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,8,8,balanced,0.3275980758666992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,8,8,balanced,0.47498622894287107
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,8,8,balanced,0.6270988845825196
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,32,2,power_law_1.2,0.3596723175048828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,8,8,balanced,0.9068134307861329
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,8,8,balanced,0.032175359725952146
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,8,8,balanced,0.03201152086257934
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,8,8,balanced,0.031765758991241455
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,8,8,balanced,0.0577459192276001
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,8,8,balanced,0.0561740779876709
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,8,8,balanced,1.210557403564453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,8,8,balanced,0.05679743766784669
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,8,8,balanced,0.057863678932189945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,8,8,balanced,0.057251839637756353
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,8,8,balanced,0.05812607765197754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,8,8,balanced,0.05821824073791504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,8,8,balanced,0.059032320976257324
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,8,8,balanced,0.05921792030334473
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,8,8,balanced,0.060371198654174806
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,8,8,balanced,0.06007808208465576
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,8,8,balanced,0.06329855918884278
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,8,8,balanced,0.06569727897644043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,8,8,balanced,0.07004672050476074
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,8,8,balanced,0.056491518020629884
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,32,2,power_law_1.2,0.273623046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,8,8,balanced,0.07421567916870117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,8,8,balanced,0.08138879776000976
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,8,8,balanced,0.07707647800445556
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,8,8,balanced,0.10053759574890138
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,8,8,balanced,0.11697664260864257
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,8,8,balanced,0.17462400436401365
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,8,8,balanced,0.20834815979003904
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,8,8,balanced,0.30024831771850585
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,8,8,power_law_1.01,0.061146879196166994
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,32,2,power_law_1.2,0.6207820892333984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,8,8,balanced,0.3849151992797851
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,8,8,power_law_1.01,0.11145088195800781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,8,8,power_law_1.01,0.1116211223602295
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,8,8,power_law_1.01,0.1128435230255127
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,8,8,power_law_1.01,0.11287296295166016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,8,8,power_law_1.01,0.11477120399475098
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,8,8,power_law_1.01,0.11685631752014161
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,8,8,power_law_1.01,0.11719679832458496
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,8,8,power_law_1.01,0.11778176307678223
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,32,2,power_law_1.2,1.0486195373535157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,8,8,power_law_1.01,0.12233471870422363
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,8,8,power_law_1.01,0.12720512390136718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,8,8,power_law_1.01,0.12713855743408203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,8,8,power_law_1.01,0.13617152214050293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,8,8,power_law_1.01,0.11099136352539063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,8,8,power_law_1.01,0.1525312042236328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,8,8,power_law_1.01,0.15873536109924316
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,8,8,power_law_1.01,0.24479999542236328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,8,8,power_law_1.01,0.12144384384155274
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,8,8,power_law_1.01,0.2896998405456543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,8,8,power_law_1.01,0.45042049407958984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,8,8,power_law_1.01,0.9274572753906251
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,8,8,power_law_1.01,0.5956915283203126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,8,8,power_law_1.01,0.03483392000198364
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,8,8,power_law_1.01,0.05638656139373779
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,8,8,power_law_1.01,0.05638912200927735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,8,8,power_law_1.01,0.05629951953887939
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,8,8,power_law_1.01,0.05642496109008789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,8,8,power_law_1.01,0.8379750061035157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,8,8,power_law_1.01,0.05714303970336913
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,8,8,power_law_1.01,0.05875967979431153
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,8,8,power_law_1.01,0.05831808090209961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,8,8,power_law_1.01,0.05981823921203613
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,8,8,power_law_1.01,0.061076478958129884
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,8,8,power_law_1.01,0.06368768215179443
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,8,8,power_law_1.01,0.06653312206268311
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,8,8,power_law_1.01,0.06886784076690675
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,8,8,power_law_1.01,0.07153791904449462
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,8,8,power_law_1.01,0.07784063816070556
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,8,8,power_law_1.01,0.07560319900512695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,8,8,power_law_1.01,0.08040831565856935
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,8,8,power_law_1.01,0.09387776374816895
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,8,8,power_law_1.01,0.1085696029663086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,8,8,power_law_1.01,1.6590821838378909
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,8,8,power_law_1.01,0.17093887329101562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,8,8,power_law_1.01,0.18415744781494142
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,8,8,power_law_1.01,0.31006336212158203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,8,8,power_law_1.01,2.215478973388672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,32,2,power_law_1.2,0.46747264862060545
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,8,8,power_law_1.01,0.28274816513061524
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,8,8,power_law_1.2,0.0612774419784546
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,8,8,power_law_1.2,0.11053055763244628
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,8,8,power_law_1.2,0.11206656455993654
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,8,8,power_law_1.2,0.11136639595031739
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,8,8,power_law_1.2,0.1121945571899414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,8,8,power_law_1.2,0.11343487739562988
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,8,8,power_law_1.2,0.11469440460205078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,8,8,power_law_1.2,0.11707648277282715
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,8,8,power_law_1.01,0.7044364929199218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,8,8,power_law_1.2,0.11684736251831054
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,8,8,power_law_1.2,0.12158335685729979
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,8,8,power_law_1.2,0.12264191627502441
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,8,8,power_law_1.2,0.12597248077392578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,8,8,power_law_1.2,0.12775296211242676
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,8,8,power_law_1.2,0.13048447608947752
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,8,8,power_law_1.01,0.5459366226196289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,8,8,power_law_1.2,0.15223679542541504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,8,8,power_law_1.2,0.17053951263427733
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,8,8,power_law_1.2,0.13731712341308594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,8,8,power_law_1.2,0.24736896514892578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,8,8,power_law_1.2,0.3106803131103516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,8,8,power_law_1.2,0.4158323287963867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,8,8,power_law_1.2,0.607459831237793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,32,2,power_law_1.2,0.6467378997802735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,8,8,power_law_1.01,1.4594598388671876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,8,8,power_law_1.2,1.1633907318115235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,8,8,power_law_1.2,0.9143052673339843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,8,8,power_law_1.2,0.031694080829620365
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,8,8,power_law_1.2,0.05743360042572021
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,8,8,power_law_1.2,0.05656320095062256
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,8,8,power_law_1.2,0.058156800270080564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,8,8,power_law_1.2,0.05702400207519531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,8,8,power_law_1.2,0.05797247886657715
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,8,8,power_law_1.2,0.058695678710937504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,8,8,power_law_1.2,0.05898367881774902
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,8,8,power_law_1.2,0.05964159965515137
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,8,8,power_law_1.2,1.7443161010742188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,8,8,power_law_1.2,0.06621439933776856
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,8,8,power_law_1.2,0.06519296169281005
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,8,8,power_law_1.2,0.06779007911682129
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,8,8,power_law_1.2,0.07006464004516602
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,8,8,power_law_1.2,0.07207039833068848
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,8,8,power_law_1.01,3.5815859985351564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,8,8,power_law_1.2,0.07565951824188233
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,8,8,power_law_1.2,0.08070015907287598
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,8,8,power_law_1.2,0.08037887573242188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,8,8,power_law_1.2,0.09867391586303711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,8,8,power_law_1.2,0.10934271812438964
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,8,8,power_law_1.2,2.253450164794922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,8,8,power_law_1.2,0.19231103897094726
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,8,8,power_law_1.2,0.17188480377197266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,8,8,power_law_1.01,4.585101928710937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,8,8,power_law_1.01,1.144711685180664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,8,8,power_law_1.2,0.5367039871215821
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,16,4,balanced,0.03631488084793091
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,16,4,balanced,0.03566720008850098
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,16,4,balanced,0.05746943950653076
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,16,4,balanced,0.10005887985229492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,16,4,balanced,0.09984895706176758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,16,4,balanced,0.10032511711120606
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,8,8,power_law_1.2,0.3760319900512695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,8,8,power_law_1.2,3.4833023071289064
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,16,4,balanced,0.101396484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,16,4,balanced,0.10118528366088868
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,16,4,balanced,0.10466943740844728
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,8,8,power_law_1.2,0.2771430397033691
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,16,4,balanced,0.10398847579956054
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,16,4,balanced,0.10532352447509766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,16,4,balanced,0.1089356803894043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,16,4,balanced,0.11091967582702637
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,16,4,balanced,0.11215999603271484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,16,4,balanced,0.11326335906982421
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,16,4,balanced,0.11819519996643066
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,16,4,balanced,0.1331993579864502
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,16,4,balanced,0.13142656326293944
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,16,4,balanced,0.1548953628540039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,16,4,balanced,0.17291135787963868
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,16,4,balanced,0.2987443161010742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,16,4,balanced,0.3449343872070313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,16,4,balanced,0.4996441650390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,16,4,balanced,0.10889599800109864
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,16,4,balanced,0.6601074981689453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,16,4,balanced,0.02874880075454712
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,16,4,balanced,0.028852479457855223
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,16,4,balanced,0.051476478576660156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,16,4,balanced,0.9535346984863281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,16,4,balanced,0.051842560768127446
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,16,4,balanced,0.05244031906127929
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,16,4,balanced,0.05240831851959229
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,16,4,balanced,0.053729281425476075
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,16,4,balanced,0.052863998413085936
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,16,4,balanced,0.05495168209075928
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,16,4,balanced,0.05492735862731933
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,16,4,balanced,0.055541758537292485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,16,4,balanced,0.055740160942077635
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,16,4,balanced,0.030982398986816408
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,16,4,balanced,0.06058368206024169
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,16,4,balanced,0.06201600074768067
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,16,4,balanced,0.0640768003463745
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,16,4,balanced,0.06892543792724609
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,16,4,balanced,0.07627136230468749
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,16,4,balanced,0.07050752162933349
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,16,4,balanced,0.08499327659606934
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,16,4,balanced,0.09433856010437011
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,16,4,balanced,0.11988479614257812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,16,4,balanced,0.14729472160339357
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,16,4,balanced,1.2621977233886719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,8,8,power_law_1.2,0.7421568298339845
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,16,4,balanced,0.22210048675537108
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,16,4,balanced,0.2662668800354004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,16,4,power_law_1.01,0.0462553596496582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,16,4,power_law_1.01,0.05387392044067383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,16,4,balanced,0.39298561096191403
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,16,4,power_law_1.01,0.09845503807067871
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,16,4,power_law_1.01,0.10005375862121582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,16,4,balanced,0.5154240036010742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,16,4,power_law_1.01,0.10024703979492186
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,16,4,power_law_1.01,0.1017843246459961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,16,4,power_law_1.01,0.1040550422668457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,16,4,power_law_1.01,0.10498687744140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,16,4,power_law_1.01,0.10804096221923828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,16,4,power_law_1.01,0.11005696296691894
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,16,4,power_law_1.01,0.11263360023498534
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,16,4,power_law_1.01,0.11403648376464844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,16,4,power_law_1.01,0.12000896453857421
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,16,4,power_law_1.01,0.12695039749145506
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,16,4,power_law_1.01,0.13560447692871094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,16,4,power_law_1.01,0.13606911659240722
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,16,4,power_law_1.01,0.16207231521606444
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,16,4,power_law_1.01,0.10799231529235839
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,16,4,power_law_1.01,0.20006784439086914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,16,4,power_law_1.01,0.30983039855957034
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,16,4,power_law_1.01,0.37243392944335935
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,16,4,power_law_1.01,0.5675033569335938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,16,4,power_law_1.01,0.7228876495361328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,16,4,power_law_1.01,0.030707199573516846
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,16,4,power_law_1.01,1.0551321411132812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,16,4,power_law_1.01,0.03242496013641357
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,16,4,power_law_1.01,0.052008957862854
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,16,4,power_law_1.01,0.051715841293334966
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,16,4,power_law_1.01,0.0521446418762207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,16,4,power_law_1.01,0.052723197937011725
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,16,4,power_law_1.01,0.05319168090820312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,16,4,power_law_1.01,0.054662399291992195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,16,4,power_law_1.01,0.054978561401367185
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,16,4,power_law_1.01,0.05626368045806884
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,16,4,power_law_1.01,0.058279681205749514
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,16,4,power_law_1.01,0.06260608196258545
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,16,4,power_law_1.01,0.06464000225067139
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,16,4,power_law_1.01,0.06779263973236084
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,16,4,power_law_1.01,0.07250432014465333
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,16,4,power_law_1.01,0.07482880115509033
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,16,4,power_law_1.01,0.07597951889038086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,16,4,power_law_1.01,0.08618240356445313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,16,4,power_law_1.01,0.093570556640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,16,4,power_law_1.01,1.3439923095703126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,16,4,power_law_1.01,0.15802111625671386
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,16,4,power_law_1.01,0.13860223770141603
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,16,4,power_law_1.01,0.23051391601562501
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,16,4,power_law_1.01,0.3928806304931641
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,16,4,power_law_1.01,0.2921126365661621
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,8,8,power_law_1.2,1.5018457031249999
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,16,4,power_law_1.2,0.04513279914855957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,16,4,power_law_1.2,0.05485695838928223
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,16,4,power_law_1.2,0.10110207557678223
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,16,4,power_law_1.2,0.10039808273315429
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,16,4,power_law_1.2,0.10067584037780761
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,16,4,power_law_1.2,0.10154111862182617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,16,4,power_law_1.2,0.10371071815490722
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,16,4,power_law_1.2,0.10510335922241212
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,16,4,power_law_1.2,0.10697728157043458
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,16,4,power_law_1.2,0.10851072311401366
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,16,4,power_law_1.2,0.11042176246643068
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,16,4,power_law_1.2,0.11238016128540038
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,16,4,power_law_1.2,0.113821439743042
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,16,4,power_law_1.2,0.12076288223266603
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,16,4,power_law_1.2,0.12734335899353028
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,16,4,power_law_1.01,0.8427597045898437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,16,4,power_law_1.2,0.1351091194152832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,16,4,power_law_1.2,0.14160256385803222
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,16,4,power_law_1.2,0.16102272033691406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,8,8,power_law_1.2,1.1579366302490235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,16,4,power_law_1.2,0.18819711685180665
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,16,4,power_law_1.2,0.3028134346008301
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,16,4,power_law_1.01,2.0050419616699218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,16,4,power_law_1.01,0.5566374588012695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,16,4,power_law_1.2,0.4065318298339844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,8,8,power_law_1.2,4.780286865234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,16,4,power_law_1.2,0.7409446716308594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,16,4,power_law_1.2,0.03047424077987671
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,16,4,power_law_1.2,0.031036159992218017
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,16,4,power_law_1.2,0.051608319282531734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,16,4,power_law_1.2,0.0517580795288086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,16,4,power_law_1.2,0.051904001235961915
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,16,4,power_law_1.2,0.05228159904479981
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,16,4,power_law_1.2,0.5352191925048828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,16,4,power_law_1.2,0.05353856086730957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,16,4,power_law_1.2,0.054302721023559576
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,16,4,power_law_1.2,0.05606527805328369
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,16,4,power_law_1.2,0.05667840003967285
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,16,4,power_law_1.2,0.059618558883666996
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,16,4,power_law_1.2,0.06279551982879639
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,16,4,power_law_1.2,0.0646131181716919
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,16,4,power_law_1.2,0.06792448043823243
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,16,4,power_law_1.2,0.07186304092407227
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,16,4,power_law_1.01,2.826871032714844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,16,4,power_law_1.2,0.07520127773284913
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,16,4,power_law_1.2,0.07510655879974365
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,16,4,power_law_1.2,0.0844927978515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,16,4,power_law_1.2,0.09769087791442871
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,16,4,power_law_1.01,1.1380313873291015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,16,4,power_law_1.2,0.14516223907470702
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,16,4,power_law_1.2,0.1699238395690918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,16,4,power_law_1.2,0.22694656372070315
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,16,4,power_law_1.2,1.0624768066406252
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,16,4,power_law_1.2,0.2984345626831055
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,32,2,balanced,0.02783744096755981
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,32,2,balanced,0.044843521118164066
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,32,2,balanced,0.055955200195312506
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,32,2,balanced,0.09477503776550293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,32,2,balanced,0.09506175994873048
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,32,2,balanced,0.09643391609191895
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,32,2,balanced,0.09777024269104004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,32,2,balanced,0.09872511863708497
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,32,2,balanced,0.09960703849792481
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,32,2,balanced,0.10034175872802735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,32,2,balanced,0.10230272293090821
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,32,2,balanced,0.10468864440917969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,32,2,balanced,0.10727680206298829
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,32,2,balanced,0.11230591773986816
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,32,2,balanced,0.11933183670043945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,32,2,balanced,0.12380288124084474
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,32,2,balanced,0.11960831642150879
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,32,2,balanced,0.13397376060485838
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,32,2,balanced,0.14660351753234863
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,32,2,balanced,0.1720332717895508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,16,4,power_law_1.2,0.44375038146972656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,32,2,balanced,0.1929190444946289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,32,2,balanced,0.32496639251708986
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,32,2,balanced,0.39202430725097653
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,16,4,power_law_1.2,1.496881866455078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,32,2,balanced,0.570458869934082
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,32,2,balanced,0.7513561248779297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,32,2,balanced,0.02750335931777954
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,32,2,balanced,1.1124864196777344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,32,2,balanced,0.029475839138031007
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,32,2,balanced,0.0325708794593811
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,32,2,balanced,0.051201281547546384
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,32,2,balanced,0.05208703994750976
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,32,2,balanced,1.4668505859374998
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,32,2,balanced,0.052033281326293944
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,32,2,balanced,0.054090237617492674
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,32,2,balanced,0.054814720153808595
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,32,2,balanced,0.05463551998138427
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,32,2,balanced,0.05779712200164795
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,32,2,balanced,0.060186882019042966
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,32,2,balanced,0.06286079883575439
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,32,2,balanced,0.05077631950378418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,32,2,balanced,0.0670527982711792
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,32,2,balanced,0.07031551837921143
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,32,2,balanced,0.07322495937347412
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,32,2,balanced,0.06821119785308838
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,32,2,balanced,0.08060671806335448
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,32,2,balanced,0.08727935791015626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,16,4,power_law_1.2,0.7651366424560547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,32,2,balanced,0.1281599998474121
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,32,2,balanced,0.17070720672607423
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,32,2,balanced,0.2106342315673828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,32,2,balanced,0.32191745758056645
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,32,2,balanced,0.40185985565185545
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,32,2,power_law_1.01,0.046417918205261234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,32,2,balanced,0.5950604629516602
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,32,2,power_law_1.01,0.05043327808380127
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,32,2,power_law_1.01,0.05194623947143555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,32,2,balanced,0.7819840240478515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,32,2,power_law_1.01,0.095731201171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,32,2,power_law_1.01,0.09644288063049317
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,32,2,power_law_1.01,0.09943424224853516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,32,2,power_law_1.01,0.09981951713562012
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,32,2,power_law_1.01,0.1008844757080078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,32,2,balanced,0.1082572841644287
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,32,2,power_law_1.01,0.09463680267333985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,32,2,power_law_1.01,0.10305279731750487
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,32,2,power_law_1.01,0.10737279891967773
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,32,2,power_law_1.01,0.11042943954467774
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,32,2,power_law_1.01,0.09790592193603516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,32,2,power_law_1.01,0.11726976394653321
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,32,2,power_law_1.01,0.12151552200317384
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,32,2,power_law_1.01,0.12357888221740723
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,32,2,power_law_1.01,0.12668928146362304
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,32,2,power_law_1.01,0.14172032356262207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,32,2,power_law_1.01,0.15509119987487793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,32,2,power_law_1.01,0.2080179214477539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,16,4,power_law_1.2,0.5695283126831054
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,32,2,power_law_1.01,0.2713548851013184
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,32,2,power_law_1.01,0.5462579345703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,32,2,power_law_1.01,0.41974910736083987
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,32,2,power_law_1.01,0.028852479457855223
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,32,2,power_law_1.01,0.030894079208374024
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,32,2,power_law_1.01,0.03404416084289551
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,32,2,power_law_1.01,0.0511897611618042
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,32,2,power_law_1.01,0.05020544052124023
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,32,2,power_law_1.01,0.05075839996337891
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,16,4,power_law_1.2,1.959710693359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,32,2,power_law_1.01,0.05302656173706055
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,32,2,power_law_1.01,0.05445888042449951
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,32,2,power_law_1.01,0.055406079292297364
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,32,2,power_law_1.01,0.056992001533508294
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,32,2,power_law_1.01,0.7484339141845704
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,32,2,power_law_1.01,0.056954879760742184
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,32,2,power_law_1.01,0.06229119777679444
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,32,2,power_law_1.01,0.0636288022994995
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,32,2,power_law_1.01,0.0697651195526123
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,32,2,power_law_1.01,0.07627776145935058
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,32,2,power_law_1.01,0.07394815921783447
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,32,2,power_law_1.01,0.07387135982513428
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,32,2,power_law_1.01,0.08421119689941406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,32,2,power_law_1.01,0.09519359588623047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,32,2,power_law_1.01,0.14159487724304198
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,32,2,power_law_1.01,0.11695743560791017
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,32,2,power_law_1.01,0.26206848144531253
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,32,2,power_law_1.01,0.20643583297729493
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,32,2,power_law_1.01,1.0251302337646484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,32,2,power_law_1.01,0.35892223358154296
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,32,2,power_law_1.2,0.04468095779418945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,32,2,power_law_1.2,0.04960127830505371
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,32,2,power_law_1.2,0.05200511932373046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,32,2,power_law_1.2,0.09499135971069336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,32,2,power_law_1.2,0.09576064109802246
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,32,2,power_law_1.2,0.0965056037902832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,32,2,power_law_1.2,0.09782655715942383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,32,2,power_law_1.2,0.09857407569885254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,32,2,power_law_1.2,0.10048000335693359
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,32,2,power_law_1.2,0.10113792419433594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,32,2,power_law_1.2,0.10393856048583985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,32,2,power_law_1.2,0.10646528244018554
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,16,4,power_law_1.2,2.5908657836914064
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,32,2,power_law_1.2,0.11235199928283693
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,32,2,power_law_1.2,0.11741312026977539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,32,2,power_law_1.2,0.12182016372680664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,32,2,power_law_1.2,0.12372223854064941
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,32,2,power_law_1.01,0.7175552368164062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,32,2,power_law_1.2,0.12695808410644532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,32,2,power_law_1.2,0.14072959899902343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,32,2,power_law_1.2,0.16435199737548828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,32,2,power_law_1.01,0.48426750183105466
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,32,2,power_law_1.2,0.2076019287109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,32,2,power_law_1.01,1.3961062622070313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,32,2,power_law_1.2,0.525291519165039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,32,2,power_law_1.2,0.26111616134643556
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,16,4,power_law_1.2,1.0291725158691407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,32,2,power_law_1.2,0.0281062388420105
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,32,2,power_law_1.2,0.03156224012374878
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,32,2,power_law_1.2,0.03354624032974243
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,32,2,power_law_1.2,0.05139711856842041
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,32,2,power_law_1.2,0.4161894226074219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,32,2,power_law_1.2,0.05132927894592285
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,32,2,power_law_1.2,0.05179264068603515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,32,2,power_law_1.2,0.05314943790435791
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,32,2,power_law_1.2,0.053217282295227056
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,32,2,power_law_1.2,0.05493504047393799
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,32,2,power_law_1.2,0.055847678184509274
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,32,2,power_law_1.2,0.05925248146057129
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,32,2,power_law_1.2,0.061054720878601074
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,32,2,power_law_1.2,0.06491648197174073
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,32,2,power_law_1.2,0.06864768028259277
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,32,2,power_law_1.2,0.746455078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,32,2,power_law_1.2,0.07382400035858154
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,32,2,power_law_1.2,0.07643136024475097
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,32,2,power_law_1.2,0.0845248031616211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,32,2,power_law_1.2,0.07191936016082764
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,32,2,power_law_1.2,0.09469183921813965
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,32,2,power_law_1.2,0.14233344078063964
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,32,2,power_law_1.2,0.1170150375366211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,32,2,power_law_1.01,1.8239436340332031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,32,2,power_law_1.2,0.260894718170166
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,32,2,power_law_1.2,0.1882598304748535
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,32,2,power_law_1.2,0.9653465270996093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,1,64,balanced,0.06796031951904297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,1,64,balanced,0.06768896102905272
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,1,64,balanced,0.06795008182525634
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,1,64,balanced,0.06791808128356933
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,1,64,balanced,0.07818367958068848
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,1,64,balanced,0.10580479621887207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,1,64,balanced,0.16489599227905274
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,1,64,balanced,0.1649113655090332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,1,64,balanced,0.16742143630981446
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,1,64,balanced,0.16751104354858398
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,1,64,balanced,0.16830720901489257
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,1,64,balanced,0.16707584381103516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,1,64,balanced,0.16765695571899414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,1,64,balanced,0.16908544540405274
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,1,64,balanced,0.17508480072021484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,1,64,balanced,0.17942527770996092
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,1,64,balanced,0.18716800689697266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,32,2,power_law_1.2,0.35683967590332033
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,1,64,balanced,0.18863359451293946
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,32,2,power_law_1.01,0.8873753356933595
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,1,64,balanced,0.2017100715637207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,1,64,balanced,0.21238399505615235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,1,64,balanced,0.25741695404052733
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,1,64,balanced,0.2869363212585449
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,1,64,balanced,0.1855449676513672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,1,64,balanced,0.4154009628295898
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,1,64,balanced,0.4765849685668945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,1,64,balanced,0.040652799606323245
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,1,64,balanced,0.6936806488037109
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,1,64,balanced,0.04015103816986084
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,1,64,balanced,0.040245761871337896
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,1,64,balanced,0.040320000648498534
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,1,64,balanced,0.8765094757080079
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,1,64,balanced,0.048266239166259765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,1,64,balanced,0.06589695930480957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,1,64,balanced,0.07324927806854248
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,1,64,balanced,0.07353600025177001
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,1,64,balanced,0.07453311920166016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,1,64,balanced,0.07450623989105225
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,1,64,balanced,0.07621632099151612
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,1,64,balanced,0.0775270414352417
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,1,64,balanced,0.07789184093475342
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,1,64,balanced,0.08069120407104492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,1,64,balanced,0.0819865608215332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,1,64,balanced,0.08408831596374512
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,1,64,balanced,0.08863231658935547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,1,64,balanced,0.09622015953063964
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,1,64,balanced,0.09234304428100586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,1,64,balanced,0.10161664009094237
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,1,64,balanced,0.1106175994873047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,1,64,balanced,0.12419584274291992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,1,64,balanced,0.14064255714416504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,1,64,balanced,0.1804915237426758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,1,64,balanced,0.2174617576599121
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,1,64,balanced,0.3033139228820801
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,1,64,power_law_1.01,0.10585984230041504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,1,64,balanced,0.3718809509277344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,1,64,power_law_1.01,0.10582655906677245
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,1,64,power_law_1.01,0.10600447654724121
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,1,64,power_law_1.01,0.1060979175567627
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,1,64,power_law_1.01,0.15731200218200683
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,1,64,power_law_1.01,0.16223743438720703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,1,64,power_law_1.01,0.1634060859680176
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,32,2,power_law_1.2,0.7245235443115234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,1,64,power_law_1.01,0.16462720870971678
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,1,64,power_law_1.01,0.16514175415039062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,32,2,power_law_1.2,0.473864974975586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,1,64,power_law_1.01,0.1699673652648926
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,1,64,power_law_1.01,0.16786943435668947
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,1,64,power_law_1.01,0.17120895385742188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,1,64,power_law_1.01,0.18105728149414063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,1,64,power_law_1.01,0.1831551933288574
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,1,64,power_law_1.01,0.18766719818115235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,1,64,power_law_1.01,0.19687423706054688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,1,64,power_law_1.01,0.21595136642456056
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,1,64,power_law_1.01,0.23917568206787107
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,1,64,power_law_1.01,0.29003263473510743
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,1,64,power_law_1.01,0.39746688842773437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,1,64,power_law_1.01,0.44387073516845704
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,1,64,power_law_1.01,0.617127685546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,1,64,power_law_1.01,0.8070630645751953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,1,64,power_law_1.01,0.0662553596496582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,1,64,power_law_1.01,1.1580761718750001
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,1,64,power_law_1.01,0.06596479892730714
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,1,64,power_law_1.01,0.06602752208709717
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,1,64,power_law_1.01,0.07031680107116699
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,1,64,power_law_1.01,0.06985472202301025
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,1,64,power_law_1.01,2.128556823730469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,1,64,power_law_1.01,0.07331456184387207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,1,64,power_law_1.01,0.07696640014648437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,1,64,power_law_1.01,0.07631616115570068
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,1,64,power_law_1.01,0.07707903861999513
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,1,64,power_law_1.01,1.8045797729492186
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,1,64,power_law_1.01,0.07973120212554932
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,1,64,power_law_1.01,0.08272383689880372
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,1,64,power_law_1.01,0.08688128471374512
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,1,64,power_law_1.01,0.09325568199157715
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,1,64,power_law_1.01,0.09644543647766113
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,1,64,power_law_1.01,0.09554688453674316
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,1,64,power_law_1.01,0.09830783843994141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,1,64,power_law_1.01,3.476282958984375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,1,64,power_law_1.01,0.11156096458435058
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,1,64,power_law_1.01,0.1263859176635742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,1,64,power_law_1.01,0.1533414363861084
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,1,64,power_law_1.01,0.21375616073608397
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,1,64,power_law_1.01,0.2748748779296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,1,64,power_law_1.01,0.39983104705810546
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,1,64,power_law_1.01,0.5635289764404297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,1,64,power_law_1.01,0.639238395690918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,1,64,power_law_1.01,0.06714623928070068
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,1,64,power_law_1.2,0.10638719558715821
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,1,64,power_law_1.2,0.1058687973022461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,1,64,power_law_1.01,0.9901529693603516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,1,64,power_law_1.2,0.1617305564880371
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,1,64,power_law_1.2,0.15786239624023438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,1,64,power_law_1.2,0.15192704200744628
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,1,64,power_law_1.2,0.16368127822875977
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,1,64,power_law_1.01,1.367362518310547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,1,64,power_law_1.2,0.1623833656311035
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,1,64,power_law_1.2,0.1673855972290039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,1,64,power_law_1.2,0.16642047882080077
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,1,64,power_law_1.2,0.1737433624267578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,1,64,power_law_1.2,0.17969152450561524
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,1,64,power_law_1.2,0.18100095748901368
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,1,64,power_law_1.2,0.18256895065307616
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,1,64,power_law_1.2,0.10586624145507813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,1,64,power_law_1.2,0.191777286529541
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,1,64,power_law_1.2,0.20121471405029298
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,1,64,power_law_1.2,0.23055744171142578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,1,64,power_law_1.2,0.29026432037353517
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,1,64,power_law_1.2,0.32669185638427733
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,1,64,power_law_1.2,0.46908542633056644
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,1,64,power_law_1.2,0.5589811325073242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,32,2,power_law_1.2,1.3860646057128907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,1,64,power_law_1.2,0.894150390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,1,64,power_law_1.2,1.0734297943115234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,1,64,power_law_1.2,0.066909441947937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,1,64,power_law_1.2,0.06634880065917968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,1,64,power_law_1.2,0.06651648044586181
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,1,64,power_law_1.2,1.5687411499023436
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,1,64,power_law_1.2,0.07228032112121582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,1,64,power_law_1.2,0.068439040184021
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,1,64,power_law_1.2,0.07202816009521484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,1,64,power_law_1.2,0.07341440200805664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,1,64,power_law_1.2,0.07533696174621582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,1,64,power_law_1.2,0.07684735774993896
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,1,64,power_law_1.2,0.08203264236450195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,1,64,power_law_1.2,2.3546981811523438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,1,64,power_law_1.2,0.08500736236572266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,1,64,power_law_1.2,0.08881024360656739
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,1,64,power_law_1.2,0.09052543640136719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,1,64,power_law_1.2,0.09242752075195312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,1,64,power_law_1.2,0.09803903579711915
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,32,2,power_law_1.2,0.9601113891601563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,1,64,power_law_1.2,0.10526464462280274
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,1,64,power_law_1.2,0.13365119934082031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,1,64,power_law_1.2,0.15218432426452638
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,1,64,power_law_1.2,0.20611072540283204
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,1,64,power_law_1.2,0.24359039306640623
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,1,64,power_law_1.2,0.07695744037628174
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,1,64,power_law_1.2,0.3709695816040039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,1,64,power_law_1.2,3.579554443359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,1,64,power_law_1.2,0.4877427291870117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,2,32,balanced,0.05567999839782715
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,2,32,balanced,0.0562227201461792
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,1,64,power_law_1.2,0.8028262329101562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,2,32,balanced,0.05647232055664062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,2,32,balanced,0.06370816230773926
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,2,32,balanced,0.07535359859466553
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,2,32,balanced,0.11295231819152832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,2,32,balanced,0.16275968551635742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,2,32,balanced,0.16310527801513672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,2,32,balanced,0.16517503738403322
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,2,32,balanced,0.16588159561157229
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,1,64,power_law_1.2,1.672410888671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,2,32,balanced,0.16879871368408203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,2,32,balanced,0.17457664489746094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,1,64,power_law_1.2,0.9079948425292969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,1,64,power_law_1.2,1.9714381408691406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,2,32,balanced,0.18451711654663086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,2,32,balanced,0.1761382484436035
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,2,32,balanced,0.18012287139892577
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,1,64,power_law_1.2,5.757205810546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,2,32,balanced,0.18640512466430664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,2,32,balanced,0.19840255737304688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,2,32,balanced,0.2165145683288574
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,2,32,balanced,0.22912256240844728
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,2,32,balanced,0.2850892829895019
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,2,32,balanced,0.31013504028320316
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,2,32,balanced,0.46392833709716796
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,2,32,balanced,0.031997439861297605
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,2,32,balanced,0.03224832057952881
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,2,32,balanced,0.03203968048095703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,2,32,balanced,0.5539801788330079
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,2,32,balanced,0.03851775884628296
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,2,32,balanced,0.1829350471496582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,2,32,balanced,0.7903705596923828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,2,32,balanced,0.178602237701416
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,2,32,balanced,0.07746175765991212
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,2,32,balanced,1.0085132598876954
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,2,32,balanced,0.07940735816955566
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,2,32,balanced,0.07918848037719727
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,2,32,balanced,0.081463041305542
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,2,32,balanced,0.08225152015686035
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,2,32,balanced,0.08371583938598633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,2,32,balanced,0.0865011215209961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,2,32,balanced,0.08966912269592285
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,2,32,balanced,0.09269760131835938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,2,32,balanced,0.0945356845855713
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,2,32,balanced,0.10224512100219726
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,2,32,balanced,0.11366911888122559
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,2,32,balanced,0.12904576301574705
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,2,32,balanced,0.15411456108093263
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,2,32,balanced,0.0484876823425293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,2,32,balanced,0.18048255920410156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,2,32,balanced,0.0683135986328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,2,32,balanced,0.24291456222534177
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,2,32,balanced,0.07689727783203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,2,32,balanced,0.29083776473999023
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,2,32,power_law_1.01,0.11490816116333007
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,2,32,power_law_1.01,0.11493760108947755
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,2,32,balanced,0.42651519775390623
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,2,32,power_law_1.01,0.11285247802734374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,2,32,power_law_1.01,0.11212672233581542
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,2,32,balanced,0.5313817596435546
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,2,32,balanced,0.0870809555053711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,2,32,power_law_1.01,0.15739904403686525
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,2,32,power_law_1.01,0.15048447608947754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,2,32,power_law_1.01,0.16412416458129883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,2,32,power_law_1.01,0.16097408294677734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,2,32,power_law_1.01,0.17194368362426757
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,2,32,power_law_1.01,0.17624704360961913
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,2,32,power_law_1.01,0.1754368019104004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,2,32,power_law_1.01,0.1829145622253418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,2,32,power_law_1.01,0.18929023742675782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,2,32,power_law_1.01,0.20105600357055664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,2,32,power_law_1.01,0.22537727355957032
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,2,32,power_law_1.01,0.2553702354431152
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,2,32,power_law_1.01,0.3123353576660156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,2,32,power_law_1.01,0.36553600311279294
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,2,32,power_law_1.01,0.16353151321411133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,2,32,power_law_1.01,0.5435084915161132
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,2,32,power_law_1.01,0.17158912658691405
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,2,32,power_law_1.01,0.715626220703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,2,32,power_law_1.01,0.06546559810638428
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,2,32,power_law_1.01,0.8956813049316406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,2,32,power_law_1.01,0.06732287883758545
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,2,32,power_law_1.01,0.06744959831237793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,2,32,power_law_1.01,0.06832640171051026
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,2,32,power_law_1.01,0.07367040157318115
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,2,32,power_law_1.01,0.0751526403427124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,2,32,power_law_1.01,1.254999008178711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,2,32,power_law_1.01,0.16961023330688477
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,2,32,power_law_1.01,0.07701632022857666
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,2,32,power_law_1.01,0.07818496227264404
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,2,32,power_law_1.01,0.08004480361938476
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,2,32,power_law_1.01,0.07893119812011719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,2,32,power_law_1.01,0.08738816261291504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,2,32,power_law_1.01,0.09105024337768555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,2,32,power_law_1.01,0.09337599754333496
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,2,32,power_law_1.01,1.7319679260253906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,2,32,power_law_1.01,0.09329792022705077
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,2,32,power_law_1.01,0.09376511573791504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,2,32,power_law_1.01,0.09659135818481446
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,2,32,power_law_1.01,0.12848640441894532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,2,32,power_law_1.01,0.11155200004577637
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,32,2,power_law_1.2,1.823747863769531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,2,32,power_law_1.01,0.1581017589569092
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,2,32,power_law_1.01,0.19378944396972658
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,2,32,power_law_1.01,2.4039385986328123
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,2,32,power_law_1.01,0.2777587127685547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,2,32,power_law_1.01,0.3452646255493164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,2,32,power_law_1.01,0.08173824310302734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,2,32,power_law_1.01,0.43696254730224615
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,2,32,power_law_1.2,0.11575936317443847
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,2,32,power_law_1.2,0.11299967765808105
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,2,32,power_law_1.2,0.11328639984130859
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,2,32,power_law_1.01,0.6297382354736328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,2,32,power_law_1.2,0.1125875186920166
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,2,32,power_law_1.2,0.16065919876098633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,2,32,power_law_1.2,0.14931967735290527
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,2,32,power_law_1.2,0.16270719528198244
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,2,32,power_law_1.2,0.1614784049987793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,2,32,power_law_1.2,0.1613644790649414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,2,32,power_law_1.2,0.1664806365966797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,2,32,power_law_1.2,0.17147775650024416
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,2,32,power_law_1.01,1.2111065673828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,2,32,power_law_1.2,0.17753087997436523
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,2,32,power_law_1.2,0.1792576026916504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,2,32,power_law_1.2,0.18083200454711915
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,2,32,power_law_1.2,0.18895488739013672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,2,32,power_law_1.2,0.20444416046142577
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,2,32,power_law_1.2,0.24657663345336914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,2,32,power_law_1.2,0.2722419166564941
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,2,32,power_law_1.2,0.37054721832275395
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,2,32,power_law_1.2,0.4265702438354492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,2,32,power_law_1.2,0.1639295959472656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,2,32,power_law_1.01,1.0126643371582031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,2,32,power_law_1.2,0.6984127807617188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,2,32,power_law_1.2,0.06541823863983154
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,2,32,power_law_1.2,0.7811225891113281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,2,32,power_law_1.2,0.07016960144042969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,2,32,power_law_1.2,0.06851967811584472
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,2,32,power_law_1.2,0.06761087894439696
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,2,32,power_law_1.2,1.1419149017333985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,2,32,power_law_1.2,0.07617792129516601
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,2,32,power_law_1.2,0.07467391967773437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,2,32,power_law_1.2,0.08073472023010254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,2,32,power_law_1.2,0.07968128204345704
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,2,32,power_law_1.2,0.08223872184753418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,2,32,power_law_1.2,0.08389887809753419
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,2,32,power_law_1.2,1.5549349975585938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,2,32,power_law_1.2,0.08584320068359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,2,32,power_law_1.2,0.09112064361572265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,2,32,power_law_1.2,0.09358847618103028
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,2,32,power_law_1.2,0.09799936294555664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,2,32,power_law_1.2,0.09618304252624513
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,2,32,power_law_1.2,0.12228480339050293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,2,32,power_law_1.2,0.13754240036010743
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,2,32,power_law_1.2,0.17764991760253906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,2,32,power_law_1.2,0.2339980888366699
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,2,32,power_law_1.2,0.3382284927368164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,2,32,power_law_1.2,0.07888512134552002
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,2,32,power_law_1.2,0.4223654556274414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,4,16,balanced,0.032065279483795166
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,2,32,power_law_1.2,0.09353983879089356
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,2,32,power_law_1.2,0.6263935852050782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,4,16,balanced,0.031823360919952394
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,4,16,balanced,0.05496064186096191
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,4,16,balanced,0.06328192234039307
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,4,16,balanced,0.07832704067230226
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,2,32,power_law_1.2,0.8363085174560547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,4,16,balanced,0.11108991622924805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,4,16,balanced,0.17395456314086916
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,4,16,balanced,0.1689945602416992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,4,16,balanced,0.1714419174194336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,4,16,balanced,0.1712460708618164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,2,32,power_law_1.2,3.467571105957031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,2,32,power_law_1.2,2.342218170166016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,4,16,balanced,0.17337087631225584
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,4,16,balanced,0.17820800781249999
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,4,16,balanced,0.18120832443237306
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,4,16,balanced,0.18338176727294922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,4,16,balanced,0.1818329620361328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,4,16,balanced,0.18886144638061525
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,4,16,balanced,0.2003545570373535
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,4,16,balanced,0.2133465576171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,4,16,balanced,0.24202367782592776
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,4,16,balanced,0.2609305572509766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,4,16,balanced,0.3349900817871094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,2,32,power_law_1.2,1.7973542785644532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,4,16,balanced,0.5887116622924805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,4,16,balanced,0.031267840862274174
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,4,16,balanced,0.03107327938079834
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,4,16,balanced,0.032240641117095944
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,4,16,balanced,0.17403263092041016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,4,16,balanced,0.7147110748291016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,2,32,power_law_1.2,1.344021759033203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,4,16,balanced,0.05180672168731689
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,4,16,balanced,0.07446784019470215
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,4,16,balanced,0.08667519569396973
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,4,16,balanced,1.0201216125488282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,4,16,balanced,0.08698495864868164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,4,16,balanced,0.08838272094726562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,4,16,balanced,0.08872320175170897
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,4,16,balanced,0.09189375877380371
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,4,16,balanced,0.09406847953796386
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,4,16,balanced,0.3837311935424805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,4,16,balanced,0.09745792388916016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,4,16,balanced,1.31844482421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,4,16,balanced,0.09028736114501953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,4,16,balanced,0.09451135635375976
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,4,16,balanced,0.09684224128723144
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,4,16,balanced,0.10257920265197754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,4,16,balanced,0.12733823776245118
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,4,16,balanced,0.17343360900878907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,4,16,balanced,0.21787776947021484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,4,16,balanced,0.2608576011657715
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,4,16,balanced,0.0389683198928833
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,4,16,balanced,0.3677734375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,4,16,power_law_1.01,0.07843967914581298
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,4,16,balanced,0.4568767929077149
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,4,16,power_law_1.01,0.10795392036437987
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,4,16,balanced,0.6758988952636719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,4,16,power_law_1.01,0.10684927940368652
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,4,16,power_law_1.01,0.1078707218170166
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,4,16,balanced,0.11452672004699707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,4,16,power_law_1.01,0.10413439750671387
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,4,16,balanced,0.14969344139099122
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,4,16,balanced,0.8617203521728516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,4,16,power_law_1.01,0.1613350486755371
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,4,16,power_law_1.01,0.1556377601623535
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,4,16,power_law_1.01,0.16082815170288084
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,4,16,power_law_1.01,0.16902143478393555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,4,16,power_law_1.01,0.17518720626831055
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,4,16,power_law_1.01,0.17726335525512696
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,4,16,power_law_1.01,0.17122432708740234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,4,16,power_law_1.01,0.1820083236694336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,4,16,power_law_1.01,0.17711360931396486
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,4,16,power_law_1.01,0.1848921585083008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,4,16,power_law_1.01,0.19305471420288084
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,4,16,power_law_1.01,0.20943103790283205
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,4,16,power_law_1.01,0.2318604850769043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,4,16,power_law_1.01,0.2960051155090332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,4,16,power_law_1.01,0.3505254364013672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,4,16,power_law_1.01,0.5086016082763672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,4,16,power_law_1.01,0.050044159889221194
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,4,16,power_law_1.01,0.6161177444458008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,4,16,power_law_1.01,0.16803327560424805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,4,16,power_law_1.01,0.06940927982330322
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,4,16,power_law_1.01,0.06948863983154296
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,4,16,power_law_1.01,0.8639730834960938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,4,16,power_law_1.01,0.06973055839538575
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,4,16,power_law_1.01,0.08108672142028808
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,4,16,power_law_1.01,0.084203519821167
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,4,16,power_law_1.01,0.08338944435119629
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,4,16,power_law_1.01,1.2064972686767579
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,4,16,power_law_1.01,0.08670207977294922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,4,16,power_law_1.01,0.09279487609863282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,4,16,power_law_1.01,0.09570816040039062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,4,16,power_law_1.01,0.09817088127136231
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,4,16,power_law_1.01,1.5505690002441406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,4,16,power_law_1.01,0.09392640113830567
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,4,16,power_law_1.01,0.09845248222351075
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,4,16,power_law_1.01,0.10664959907531739
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,4,16,power_law_1.01,0.12032896041870118
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,4,16,power_law_1.01,0.06972415924072266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,4,16,power_law_1.01,0.17413631439208985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,4,16,power_law_1.01,0.2151910400390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,4,16,power_law_1.01,0.30601728439331055
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,4,16,power_law_1.01,2.353729248046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,4,16,power_law_1.01,0.08743680000305176
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,4,16,power_law_1.01,0.3978099060058594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,4,16,power_law_1.01,0.09182463645935059
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,4,16,power_law_1.01,0.5133555221557617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,4,16,power_law_1.2,0.10659199714660646
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,4,16,power_law_1.2,0.09868288040161133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,4,16,power_law_1.01,0.13938816070556642
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,4,16,power_law_1.2,0.10816384315490724
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,4,16,power_law_1.2,0.10775296211242676
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,4,16,power_law_1.2,0.15127039909362794
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,4,16,power_law_1.01,1.0121024322509764
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,4,16,power_law_1.2,0.16454015731811525
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,4,16,power_law_1.2,0.16868352890014648
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,4,16,power_law_1.2,0.1683135986328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,4,16,power_law_1.2,0.1765836715698242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,4,16,power_law_1.2,0.17815296173095702
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,4,16,power_law_1.01,1.4376243591308593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,4,16,power_law_1.2,0.17727872848510742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,4,16,power_law_1.2,0.1785945510864258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,4,16,power_law_1.2,0.19198720932006835
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,4,16,power_law_1.01,0.662077407836914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,4,16,power_law_1.2,0.19911680221557618
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,4,16,power_law_1.2,0.22372095108032228
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,4,16,power_law_1.2,0.2600435256958008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,4,16,power_law_1.2,0.3509260940551758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,4,16,power_law_1.2,0.1594431972503662
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,4,16,power_law_1.2,0.39706878662109374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,4,16,power_law_1.2,0.0790502405166626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,4,16,power_law_1.2,0.6217612838745117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,4,16,power_law_1.2,0.7609600067138672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,4,16,power_law_1.2,0.05016320228576661
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,4,16,power_law_1.2,0.06834176063537598
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,4,16,power_law_1.2,1.0597414398193359
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,4,16,power_law_1.2,0.0640447998046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,4,16,power_law_1.2,0.07252607822418214
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,4,16,power_law_1.2,0.07991680145263672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,4,16,power_law_1.2,1.3886528015136719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,4,16,power_law_1.2,0.17163135528564452
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,4,16,power_law_1.2,0.08615039825439454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,4,16,power_law_1.2,0.08617728233337403
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,4,16,power_law_1.2,0.08747520446777343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,4,16,power_law_1.2,0.08779135704040528
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,4,16,power_law_1.2,1.907457275390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,4,16,power_law_1.2,0.09157631874084474
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,4,16,power_law_1.2,0.09325183868408203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,4,16,power_law_1.2,0.09127296447753906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,4,16,power_law_1.2,0.09343104362487793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,4,16,power_law_1.2,0.10114432334899903
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,4,16,power_law_1.2,0.09584383964538574
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,4,16,power_law_1.2,0.10846591949462892
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,4,16,power_law_1.2,0.1337932777404785
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,4,16,power_law_1.2,0.1558348846435547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,4,16,power_law_1.2,0.07136896133422851
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,4,16,power_law_1.2,0.22364927291870118
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,4,16,power_law_1.2,0.3244966506958008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,4,16,power_law_1.2,2.665593566894531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,4,16,power_law_1.2,0.4253196716308594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,8,8,balanced,0.037482240200042725
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,4,16,power_law_1.2,0.5913318252563476
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,8,8,balanced,0.031744000911712644
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,8,8,balanced,0.06658559799194336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,8,8,balanced,0.08045184135437011
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,8,8,balanced,0.116746244430542
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,4,16,power_law_1.2,0.906160659790039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,8,8,balanced,0.1705036735534668
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,4,16,power_law_1.2,0.19942144393920896
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,8,8,balanced,0.177391357421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,8,8,balanced,0.1799590492248535
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,8,8,balanced,0.17278079986572265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,4,16,power_law_1.2,1.1923865509033202
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,8,8,balanced,0.17618688583374023
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,8,8,balanced,0.18044288635253908
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,8,8,balanced,0.18679807662963868
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,8,8,balanced,0.19304191589355468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,8,8,balanced,0.2045516777038574
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,8,8,balanced,0.22573312759399414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,8,8,balanced,0.24873472213745118
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,8,8,balanced,0.055774722099304196
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,8,8,balanced,0.2926144027709961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,8,8,balanced,0.3343206405639648
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,8,8,balanced,0.4507904052734375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,8,8,balanced,0.17341184616088867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,8,8,balanced,0.5404927825927734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,8,8,balanced,0.03142656087875366
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,8,8,balanced,0.03193984031677246
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,8,8,balanced,0.8429376220703124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,8,8,balanced,0.035238399505615234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,8,8,balanced,0.19651071548461913
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,8,8,balanced,0.04451456069946289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,8,8,balanced,1.0493465423583985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,8,8,balanced,0.05860095977783203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,8,8,balanced,0.10174592018127442
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,8,8,balanced,0.10435327529907226
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,8,8,balanced,0.11078656196594239
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,8,8,balanced,0.10828672409057619
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,8,8,balanced,0.10316800117492675
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,8,8,balanced,0.10586624145507813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,8,8,balanced,0.1081049633026123
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,8,8,balanced,0.11438207626342774
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,8,8,balanced,0.11865728378295899
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,8,8,balanced,0.12318079948425294
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,8,8,balanced,0.13431167602539062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,8,8,balanced,0.15730815887451172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,8,8,balanced,0.17913856506347656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,8,8,balanced,0.21863168716430664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,8,8,balanced,0.26064640045166015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,8,8,balanced,1.9313880920410156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,8,8,balanced,0.3445183944702149
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,4,16,power_law_1.2,1.6879014587402346
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,8,8,balanced,0.4264422225952148
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,8,8,power_law_1.01,0.06784128189086915
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,8,8,power_law_1.01,0.07715199947357178
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,8,8,balanced,1.4898725891113282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,8,8,balanced,0.7871231842041015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,8,8,power_law_1.01,0.09036671638488769
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,8,8,power_law_1.01,0.0907481575012207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,8,8,power_law_1.01,0.10107904434204103
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,8,8,power_law_1.01,0.16272127151489257
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,8,8,power_law_1.01,0.10477824211120605
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,8,8,balanced,1.1731724548339844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,8,8,balanced,0.08510592460632324
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,8,8,power_law_1.01,0.16236032485961913
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,8,8,power_law_1.01,0.16593151092529296
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,8,8,power_law_1.01,0.17278976440429689
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,8,8,power_law_1.01,0.17266304016113282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,8,8,power_law_1.01,0.17014144897460937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,8,8,power_law_1.01,0.17291519165039063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,8,8,balanced,1.5254348754882812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,8,8,balanced,0.623480339050293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,8,8,power_law_1.01,0.17490047454833985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,8,8,power_law_1.01,0.17906560897827148
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,8,8,power_law_1.01,0.19067392349243165
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,8,8,power_law_1.01,0.2096563148498535
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,8,8,power_law_1.01,0.2379136085510254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,8,8,power_law_1.01,0.27876991271972656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,8,8,power_law_1.01,0.34402305603027344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,8,8,power_law_1.01,0.1600192070007324
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,8,8,power_law_1.01,0.4488383865356445
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,8,8,power_law_1.01,0.5798105621337891
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,8,8,power_law_1.01,0.041326079368591305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,8,8,power_law_1.01,0.05497087955474853
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,8,8,power_law_1.01,0.8356607818603516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,8,8,power_law_1.01,0.06459263801574708
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,8,8,power_law_1.01,0.06687104225158691
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,8,8,power_law_1.01,0.07263872146606445
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,8,8,power_law_1.01,0.07020031929016113
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,8,8,power_law_1.01,0.09541376113891602
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,8,8,power_law_1.01,0.09785856246948242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,8,8,power_law_1.01,0.10043904304504395
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,8,8,power_law_1.01,0.10632960319519043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,8,8,power_law_1.01,1.5320729064941407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,8,8,power_law_1.01,0.10583807945251464
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,8,8,power_law_1.01,0.10609536170959473
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,8,8,power_law_1.01,0.10669823646545411
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,8,8,power_law_1.01,0.10743424415588379
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,8,8,power_law_1.01,0.11322879791259766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,8,8,power_law_1.01,0.12419967651367188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,8,8,power_law_1.01,0.15206144332885743
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,8,8,power_law_1.01,0.16610048294067384
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,8,8,power_law_1.01,0.20274431228637696
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,8,8,power_law_1.01,1.133975067138672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,8,8,power_law_1.01,0.09520383834838866
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,8,8,power_law_1.01,0.24818687438964843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,8,8,power_law_1.01,0.30070528030395505
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,8,8,power_law_1.01,0.4221324920654297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,8,8,power_law_1.2,0.06423808097839355
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,8,8,power_law_1.2,0.07683712005615234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,8,8,power_law_1.01,0.5847359848022461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,8,8,power_law_1.2,0.07521408081054688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,8,8,power_law_1.2,0.08944512367248535
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,8,8,power_law_1.2,0.10182399749755859
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,8,8,power_law_1.01,0.7755686187744141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,8,8,power_law_1.2,0.1572211170196533
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,8,8,power_law_1.2,0.10142208099365235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,8,8,power_law_1.2,0.1617215919494629
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,8,8,power_law_1.01,2.1254975891113284
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,8,8,power_law_1.01,1.0906278228759765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,8,8,power_law_1.2,0.1717158317565918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,8,8,power_law_1.2,0.16625024795532226
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,8,8,power_law_1.2,0.16648448944091795
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,8,8,power_law_1.2,0.1761907196044922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,8,8,power_law_1.2,0.18078079223632812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,8,8,power_law_1.01,1.4732940673828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,8,8,power_law_1.2,0.19924095153808594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,8,8,power_law_1.2,0.21448320388793946
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,8,8,power_law_1.2,0.2404582405090332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,8,8,power_law_1.2,0.3091302490234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,8,8,power_law_1.2,0.36243457794189454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,8,8,power_law_1.2,0.49240447998046877
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,8,8,power_law_1.2,0.16427135467529297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,8,8,power_law_1.2,0.5795865631103515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,8,8,power_law_1.2,0.1691276741027832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,8,8,power_law_1.2,0.04176767826080322
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,8,8,power_law_1.2,0.0543782377243042
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,8,8,power_law_1.2,0.05733888149261475
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,8,8,power_law_1.2,0.9299686431884766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,8,8,power_law_1.2,0.0655897617340088
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,8,8,power_law_1.2,0.07282559871673584
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,8,8,power_law_1.2,0.07322368144989014
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,8,8,power_law_1.2,1.170429458618164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,8,8,power_law_1.2,0.09428095817565918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,8,8,power_law_1.2,0.09548159599304198
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,8,8,power_law_1.2,0.09744383811950683
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,8,8,power_law_1.2,0.1024499225616455
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,8,8,power_law_1.2,0.10377471923828126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,8,8,power_law_1.2,0.10039808273315429
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,8,8,power_law_1.2,0.10862463951110839
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,8,8,power_law_1.2,0.11119487762451172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,8,8,power_law_1.2,0.11484160423278808
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,8,8,power_law_1.2,0.12459903717041017
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,8,8,power_law_1.2,0.1835148811340332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,8,8,power_law_1.2,1.7859954833984375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,8,8,power_law_1.2,0.14859392166137694
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,8,8,power_law_1.2,0.16948863983154297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,8,8,power_law_1.2,0.21990144729614255
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,8,8,power_law_1.2,0.256496639251709
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,8,8,power_law_1.2,0.3344985580444336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,8,8,power_law_1.2,2.3456460571289064
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,16,4,balanced,0.030795519351959226
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,8,8,power_law_1.2,0.10760319709777832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,8,8,power_law_1.2,0.5999436950683593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,16,4,balanced,0.058018560409545894
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,16,4,balanced,0.050895361900329596
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,16,4,balanced,0.06994431972503662
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,8,8,power_law_1.2,0.8291494750976562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,16,4,balanced,0.11842816352844239
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,16,4,balanced,0.11907072067260742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,16,4,balanced,0.1222771167755127
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,16,4,balanced,0.1702297592163086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,16,4,balanced,0.16846464157104493
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,16,4,balanced,0.17347455978393556
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,16,4,balanced,0.17625343322753906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,16,4,balanced,0.1786675262451172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,8,8,power_law_1.2,1.2491302490234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,16,4,balanced,0.18462080001831055
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,8,8,power_law_1.2,0.4167308807373047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,16,4,balanced,0.19421440124511719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,16,4,balanced,0.2030975914001465
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,16,4,balanced,0.2221887969970703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,8,8,power_law_1.2,1.7304025268554688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,16,4,balanced,0.24175487518310548
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,16,4,balanced,0.2835238456726074
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,16,4,balanced,0.08721407890319824
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,16,4,balanced,0.32699390411376955
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,16,4,balanced,0.4374540710449219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,16,4,balanced,0.03053567886352539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,16,4,balanced,0.523502082824707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,16,4,balanced,0.03237504005432129
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,16,4,balanced,0.18959232330322268
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,16,4,balanced,0.03656831979751587
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,16,4,balanced,0.8192781066894531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,16,4,balanced,0.05043968200683594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,16,4,balanced,1.0886771392822265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,16,4,balanced,0.09788928031921387
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,16,4,balanced,0.09962752342224121
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,16,4,balanced,0.1013145637512207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,16,4,balanced,0.12855423927307127
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,16,4,balanced,0.12750080108642578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,16,4,balanced,0.13143936157226563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,16,4,balanced,0.1341196823120117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,16,4,balanced,1.5301913452148437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,16,4,balanced,0.13583744049072266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,16,4,balanced,0.14020223617553712
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,16,4,balanced,0.14510463714599608
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,16,4,balanced,0.1490291213989258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,16,4,balanced,0.1785638427734375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,16,4,balanced,0.19875200271606447
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,16,4,balanced,0.23961984634399416
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,16,4,balanced,0.2785049629211426
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,16,4,balanced,2.0220249938964843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,16,4,balanced,0.06710527896881104
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,16,4,balanced,0.3560076904296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,16,4,balanced,0.43355262756347657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,16,4,power_law_1.01,0.05524991989135743
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,16,4,power_law_1.01,0.06440063953399658
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,16,4,balanced,0.6495565032958984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,16,4,power_law_1.01,0.06871935844421387
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,16,4,power_law_1.01,0.07965568065643311
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,16,4,power_law_1.01,0.08707072257995604
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,16,4,balanced,0.8093119812011718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,16,4,balanced,0.15809663772583007
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,16,4,power_law_1.01,0.10935040473937989
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,16,4,power_law_1.01,0.11235072135925292
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,16,4,power_law_1.01,0.15895808219909668
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,16,4,power_law_1.01,0.162609920501709
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,16,4,power_law_1.01,0.1660505676269531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,16,4,power_law_1.01,0.1689548873901367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,16,4,power_law_1.01,0.17519359588623046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,16,4,power_law_1.01,0.17926912307739257
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,16,4,power_law_1.01,0.1833292770385742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,16,4,power_law_1.01,0.19676927566528318
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,16,4,power_law_1.01,0.22002304077148435
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,16,4,power_law_1.01,0.24449279785156247
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,16,4,power_law_1.01,0.2888819122314453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,16,4,power_law_1.01,0.0907532787322998
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,16,4,power_law_1.01,0.34068607330322265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,16,4,power_law_1.01,0.45414142608642577
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,16,4,power_law_1.01,0.15780608177185057
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,16,4,balanced,1.2092505645751952
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,16,4,power_law_1.01,0.5799091339111329
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,16,4,power_law_1.01,0.03492480039596558
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,16,4,power_law_1.01,0.04628352165222168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,16,4,balanced,1.5804301452636718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,16,4,power_law_1.01,0.0564467191696167
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,16,4,power_law_1.01,0.8833856201171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,16,4,power_law_1.01,0.06236671924591064
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,16,4,power_law_1.01,0.06811903953552247
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,16,4,power_law_1.01,0.07317376136779785
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,16,4,power_law_1.01,1.1341503906250001
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,16,4,power_law_1.01,0.09305215835571289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,16,4,power_law_1.01,0.11552512168884277
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,16,4,power_law_1.01,0.11702400207519532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,16,4,power_law_1.01,0.12111871719360352
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,16,4,power_law_1.01,0.12534144401550293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,16,4,power_law_1.01,0.1273356819152832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,16,4,power_law_1.01,0.13423999786376953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,16,4,power_law_1.01,1.5853414916992188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,16,4,power_law_1.01,0.13586688041687012
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,16,4,power_law_1.01,0.14027647972106932
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,16,4,power_law_1.01,0.1496115207672119
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,16,4,power_law_1.01,0.17423999786376954
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,16,4,power_law_1.01,0.19447679519653321
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,16,4,power_law_1.01,0.2431577682495117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,16,4,power_law_1.01,0.29302783966064455
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,16,4,power_law_1.01,0.38239742279052735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,16,4,power_law_1.01,0.08668543815612792
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,16,4,power_law_1.01,0.4695718383789062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,16,4,power_law_1.2,0.05477248191833496
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,16,4,power_law_1.2,0.06446847915649415
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,16,4,power_law_1.01,0.69185791015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,16,4,power_law_1.2,0.06809088230133056
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,16,4,power_law_1.2,0.07905663967132567
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,16,4,power_law_1.2,0.08353664398193358
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,16,4,power_law_1.2,0.09713536262512207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,16,4,power_law_1.2,0.10909312248229981
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,16,4,power_law_1.01,1.2520883178710938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,16,4,power_law_1.2,0.11086336135864258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,16,4,power_law_1.01,2.07891845703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,16,4,power_law_1.2,0.1556991958618164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,16,4,power_law_1.2,0.15759615898132323
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,16,4,power_law_1.01,1.5957965087890624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,16,4,power_law_1.2,0.16018560409545898
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,16,4,power_law_1.2,0.169117431640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,16,4,power_law_1.2,0.17560319900512694
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,16,4,power_law_1.2,0.1802176094055176
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,16,4,power_law_1.2,0.20144639968872072
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,16,4,power_law_1.01,0.8825190734863281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,16,4,power_law_1.2,0.22524927139282225
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,16,4,power_law_1.2,0.24627967834472658
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,16,4,power_law_1.2,0.29648000717163087
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,16,4,power_law_1.2,0.3655526351928711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,16,4,power_law_1.2,0.5024076843261719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,16,4,power_law_1.2,0.6286963272094727
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,16,4,power_law_1.2,0.03535360097885132
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,16,4,power_law_1.2,0.8893235015869141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,16,4,power_law_1.2,0.18878591537475586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,16,4,power_law_1.2,0.04647808074951172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,16,4,power_law_1.2,1.1635250854492187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,16,4,power_law_1.2,0.05138048171997071
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,16,4,power_law_1.2,0.06123007774353027
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,16,4,power_law_1.2,0.07130112171173095
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,16,4,power_law_1.2,0.07376895904541016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,16,4,power_law_1.2,0.09041407585144043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,16,4,power_law_1.2,0.11523327827453614
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,16,4,power_law_1.2,0.11812479972839356
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,16,4,power_law_1.2,0.16132863998413086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,16,4,power_law_1.2,0.11887359619140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,16,4,power_law_1.2,0.12521087646484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,16,4,power_law_1.2,0.13508864402770998
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,16,4,power_law_1.2,1.7141952514648438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,16,4,power_law_1.2,0.135796480178833
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,16,4,power_law_1.2,0.14359040260314943
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,16,4,power_law_1.2,0.15279104232788085
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,16,4,power_law_1.2,0.17907455444335937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,16,4,power_law_1.2,0.199552001953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,16,4,power_law_1.2,2.145786895751953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,16,4,power_law_1.2,0.304531192779541
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,16,4,power_law_1.2,0.08890879631042481
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,16,4,power_law_1.2,0.40766975402832034
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,16,4,power_law_1.2,0.5205376052856445
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,32,2,balanced,0.03150207996368408
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,16,4,power_law_1.2,0.12482687950134277
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,32,2,balanced,0.04387968063354492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,32,2,balanced,0.05946368217468262
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,16,4,power_law_1.2,0.7157810974121094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,32,2,balanced,0.0739686393737793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,32,2,balanced,0.0997811222076416
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,32,2,balanced,0.1390169620513916
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,32,2,balanced,0.14011520385742188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,16,4,power_law_1.2,0.24493568420410156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,32,2,balanced,0.14293888092041016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,16,4,power_law_1.2,0.9011481475830078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,32,2,balanced,0.14801664352416993
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,32,2,balanced,0.18750463485717775
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,32,2,balanced,0.19020799636840818
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,32,2,balanced,0.19887104034423828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,16,4,power_law_1.2,1.2995429992675782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,32,2,balanced,0.20164096832275388
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,32,2,balanced,0.21083263397216795
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,32,2,balanced,0.22927360534667968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,32,2,balanced,0.248852481842041
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,16,4,power_law_1.2,1.7026010131835938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,32,2,balanced,0.33238014221191403
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,32,2,balanced,0.14110591888427734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,32,2,balanced,0.4389363098144531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,32,2,balanced,0.14423168182373047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,32,2,balanced,0.5247820663452148
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,32,2,balanced,0.031690239906311035
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,32,2,balanced,0.1941747283935547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,32,2,balanced,0.03514240026473999
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,32,2,balanced,0.04234496116638183
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,32,2,balanced,0.815572509765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,32,2,balanced,0.05916800022125244
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,32,2,balanced,0.09109375953674317
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,32,2,balanced,0.14134143829345702
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,32,2,balanced,0.1429964828491211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,32,2,balanced,0.2902873611450195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,32,2,balanced,0.14734975814819334
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,32,2,balanced,0.14721535682678222
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,32,2,balanced,0.1497049617767334
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,32,2,balanced,0.1984332847595215
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,32,2,balanced,0.20063871383666992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,32,2,balanced,0.2037593650817871
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,32,2,balanced,0.20843008041381833
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,32,2,balanced,1.6187776184082032
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,32,2,balanced,0.22095232009887694
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,32,2,balanced,0.23623552322387695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,32,2,balanced,0.25278207778930667
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,32,2,balanced,0.29168512344360353
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,32,2,balanced,0.14447615623474122
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,32,2,balanced,1.0847769927978514
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,32,2,balanced,2.175225524902344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,32,2,balanced,0.40164096832275387
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,32,2,balanced,0.47742847442626957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,32,2,balanced,0.21137792587280274
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,32,2,power_law_1.01,0.03318527936935425
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,32,2,balanced,0.7685568237304687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,32,2,power_law_1.01,0.05661183834075928
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,32,2,balanced,0.9228684997558594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,32,2,balanced,0.3268108749389648
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,32,2,power_law_1.01,0.07705088138580322
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,32,2,power_law_1.01,0.0949772834777832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,32,2,power_law_1.01,0.1188697624206543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,32,2,power_law_1.01,0.10591103553771972
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,32,2,balanced,1.3604301452636718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,32,2,power_law_1.01,0.12473088264465333
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,32,2,power_law_1.01,0.12730112075805664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,32,2,power_law_1.01,0.1307699203491211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,32,2,power_law_1.01,0.13870464324951173
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,32,2,power_law_1.01,0.17363584518432618
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,32,2,power_law_1.01,0.17765247344970703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,32,2,power_law_1.01,0.18453887939453123
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,32,2,balanced,1.8157452392578126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,32,2,power_law_1.01,0.18856063842773438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,32,2,power_law_1.01,0.19366399765014647
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,32,2,power_law_1.01,0.2048588752746582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,32,2,power_law_1.01,0.06702847957611084
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,32,2,power_law_1.01,0.25297664642333983
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,32,2,power_law_1.01,0.3006080055236816
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,32,2,power_law_1.01,0.35005184173583986
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,32,2,power_law_1.01,0.44814590454101566
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,32,2,power_law_1.01,0.5877695846557617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,32,2,power_law_1.01,0.03462143898010254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,32,2,power_law_1.01,0.040965118408203126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,32,2,power_law_1.01,0.8704192352294922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,32,2,power_law_1.01,0.05328256130218506
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,32,2,power_law_1.01,0.06689663887023925
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,32,2,power_law_1.01,0.08288384437561035
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,32,2,power_law_1.01,0.23150207519531252
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,32,2,power_law_1.01,1.1233856201171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,32,2,power_law_1.01,0.11578880310058595
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,32,2,power_law_1.01,0.120065279006958
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,32,2,power_law_1.01,0.12488063812255859
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,32,2,power_law_1.01,0.12709247589111328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,32,2,power_law_1.01,0.1327577590942383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,32,2,power_law_1.01,0.17685632705688475
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,32,2,power_law_1.01,0.1814259147644043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,32,2,power_law_1.01,1.569027862548828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,32,2,power_law_1.01,0.19079423904418946
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,32,2,power_law_1.01,0.1863705635070801
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,32,2,power_law_1.01,0.19180543899536134
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,32,2,power_law_1.01,0.2080089569091797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,32,2,power_law_1.01,0.23183872222900392
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,32,2,power_law_1.01,0.2535372734069824
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,32,2,power_law_1.01,0.3406016159057617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,32,2,power_law_1.01,0.3962521743774414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,32,2,power_law_1.01,2.121442565917969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,32,2,power_law_1.01,0.09649408340454102
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,32,2,power_law_1.01,0.496447982788086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,32,2,power_law_1.01,0.6024448013305663
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,32,2,power_law_1.2,0.033397760391235355
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,32,2,power_law_1.2,0.057121281623840336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,32,2,power_law_1.2,0.06443007946014404
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,32,2,power_law_1.01,0.8399091339111328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,32,2,power_law_1.2,0.07623936176300049
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,32,2,power_law_1.2,0.09381119728088379
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,32,2,power_law_1.2,0.10280192375183106
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,32,2,power_law_1.2,0.12195584297180176
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,32,2,power_law_1.01,1.0396723175048828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,32,2,power_law_1.2,0.1280998420715332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,32,2,power_law_1.2,0.13172863960266112
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,32,2,power_law_1.2,0.13828096389770508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,32,2,power_law_1.2,0.17588096618652344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,32,2,power_law_1.2,0.18458751678466795
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,32,2,power_law_1.2,0.19078655242919923
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,32,2,power_law_1.2,0.19510528564453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,32,2,power_law_1.2,0.20656896591186524
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,32,2,power_law_1.01,1.525745849609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,32,2,power_law_1.2,0.23500288009643552
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,32,2,power_law_1.2,0.258922233581543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,32,2,power_law_1.2,0.3087718391418457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,32,2,power_law_1.2,0.35937408447265623
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,32,2,power_law_1.01,1.9492095947265624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,32,2,power_law_1.2,0.4584294509887695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,32,2,power_law_1.2,0.12265600204467772
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,32,2,power_law_1.2,0.03506815910339355
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,32,2,power_law_1.2,0.6005875015258789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,32,2,power_law_1.2,0.041300477981567385
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,32,2,power_law_1.2,0.17799936294555666
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,32,2,power_law_1.2,0.8922560119628906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,32,2,power_law_1.2,0.06526463985443115
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,32,2,power_law_1.2,0.07963776111602783
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,32,2,power_law_1.2,0.09437952041625977
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,32,2,power_law_1.2,1.1339186859130859
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,32,2,power_law_1.2,0.11920512199401856
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,32,2,power_law_1.2,0.11991168022155763
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,32,2,power_law_1.2,0.12348031997680664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,32,2,power_law_1.2,0.12629504203796388
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,32,2,power_law_1.2,0.1318617630004883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,32,2,power_law_1.2,0.17834367752075195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,32,2,power_law_1.2,0.18173696517944335
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,32,2,power_law_1.2,0.1899635124206543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,32,2,power_law_1.2,1.626946563720703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,32,2,power_law_1.2,0.19108991622924804
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,32,2,power_law_1.2,0.1981376075744629
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,32,2,power_law_1.2,0.21343488693237306
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,32,2,power_law_1.2,0.05028223991394043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,32,2,power_law_1.2,0.2499519920349121
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,32,2,power_law_1.2,0.29383039474487305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,32,2,power_law_1.2,0.4088307189941406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,32,2,power_law_1.2,0.5130022430419923
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,32,2,power_law_1.2,0.6132556915283203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,1,64,balanced,0.061770238876342774
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,32,2,power_law_1.2,0.8608614349365235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,1,64,balanced,0.06163584232330323
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,1,64,balanced,0.061643519401550294
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,1,64,balanced,0.06189184188842774
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,32,2,power_law_1.2,1.0763225555419922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,32,2,power_law_1.2,0.35597694396972657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,1,64,balanced,0.07088128089904785
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,1,64,balanced,0.09406463623046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,1,64,balanced,0.1475494384765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,1,64,balanced,0.1490713596343994
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,1,64,balanced,0.15147135734558106
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,1,64,balanced,0.1515174388885498
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,32,2,power_law_1.2,1.5619775390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,1,64,balanced,0.1492915153503418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,1,64,balanced,0.14961536407470705
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,1,64,balanced,0.15675264358520508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,1,64,balanced,0.15926783561706542
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,1,64,balanced,0.16572032928466798
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,1,64,balanced,0.16656383514404297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,1,64,balanced,0.16982656478881836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,1,64,balanced,0.1790835189819336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,1,64,balanced,0.1849203109741211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,32,2,power_law_1.2,2.011700439453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,32,2,power_law_1.2,2.170714874267578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,1,64,balanced,0.22798463821411136
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,1,64,balanced,0.24507904052734375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,1,64,balanced,0.1463257598876953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,1,64,balanced,0.036595199108123776
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,1,64,balanced,0.03581696033477783
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,1,64,balanced,0.3641856002807617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,1,64,balanced,0.4174694442749024
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,1,64,balanced,0.03537152051925659
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,1,64,balanced,0.04496255874633789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,1,64,balanced,0.059951357841491706
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,1,64,balanced,0.06640128135681153
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,1,64,balanced,0.5994611358642578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,1,64,balanced,0.06594431877136231
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,1,64,balanced,0.06897664070129395
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,1,64,balanced,0.7476121520996093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,1,64,balanced,0.06695295810699463
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,1,64,balanced,0.06844672203063965
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,1,64,balanced,0.07022592067718506
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,1,64,balanced,0.07016064167022705
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,1,64,balanced,0.07814271926879883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,1,64,balanced,0.0762828779220581
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,1,64,balanced,0.07919871807098389
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,1,64,balanced,0.08826496124267577
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,1,64,balanced,0.08239744186401368
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,1,64,balanced,0.09095935821533203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,1,64,balanced,0.09972352027893067
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,1,64,balanced,0.11421312332153319
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,1,64,balanced,0.1283724784851074
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,1,64,balanced,0.14840831756591796
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,1,64,balanced,0.197126407623291
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,1,64,balanced,0.276177921295166
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,1,64,balanced,0.036542720794677734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,1,64,power_law_1.01,0.09434623718261718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,1,64,balanced,0.3378905487060547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,1,64,power_law_1.01,0.09448575973510742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,1,64,power_law_1.01,0.09427583694458008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,1,64,power_law_1.01,0.13326848030090332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,1,64,power_law_1.01,0.14500224113464355
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,1,64,power_law_1.01,0.14898048400878908
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,1,64,balanced,0.07475456237792968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,1,64,power_law_1.01,0.1509990406036377
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,1,64,power_law_1.01,0.1499724769592285
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,1,64,power_law_1.01,0.15086591720581055
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,1,64,power_law_1.01,0.15734656333923341
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,1,64,power_law_1.01,0.15849216461181642
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,1,64,power_law_1.01,0.1599500846862793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,1,64,balanced,0.1659174346923828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,1,64,power_law_1.01,0.16969343185424807
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,1,64,power_law_1.01,0.17659648895263672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,1,64,power_law_1.01,0.18539264678955078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,1,64,power_law_1.01,0.22120832443237304
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,1,64,power_law_1.01,0.250384635925293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,1,64,power_law_1.01,0.0937446403503418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,1,64,power_law_1.01,0.35336063385009764
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,1,64,power_law_1.01,0.3934988784790039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,1,64,power_law_1.01,0.6019942474365234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,1,64,power_law_1.01,0.06039167881011963
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,1,64,power_law_1.01,0.6926783752441407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,1,64,power_law_1.01,0.060476160049438475
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,1,64,power_law_1.01,0.16548992156982423
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,1,64,power_law_1.01,0.06019839763641357
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,1,64,power_law_1.01,0.9797567749023438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,1,64,power_law_1.01,0.06054272174835205
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,1,64,power_law_1.01,0.06219520092010498
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,1,64,power_law_1.01,0.06518271923065186
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,1,64,power_law_1.01,0.06719232082366944
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,1,64,power_law_1.01,0.0691596794128418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,1,64,power_law_1.01,0.06938240051269531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,1,64,power_law_1.01,0.0714572811126709
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,1,64,power_law_1.01,0.07423359870910644
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,1,64,power_law_1.01,0.07751552104949952
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,1,64,power_law_1.01,0.07896319866180421
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,1,64,power_law_1.01,0.08310784339904785
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,1,64,power_law_1.01,0.08674943923950196
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,1,64,power_law_1.01,0.08707712173461915
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,1,64,power_law_1.01,0.09046527862548828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,1,64,power_law_1.01,0.10322303771972656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,1,64,power_law_1.01,0.11277183532714843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,1,64,power_law_1.01,2.1531558227539063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,1,64,power_law_1.01,0.15161343574523928
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,1,64,power_law_1.01,0.1915532875061035
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,1,64,power_law_1.01,0.2769152069091797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,1,64,power_law_1.01,0.33359233856201176
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,1,64,power_law_1.01,1.4351065063476562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,1,64,power_law_1.01,0.41342079162597656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,1,64,power_law_1.2,0.09443967819213868
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,1,64,power_law_1.01,2.8352728271484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,1,64,power_law_1.01,0.6197951889038086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,1,64,power_law_1.2,0.09335040092468262
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,1,64,power_law_1.2,0.09439871788024903
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,1,64,power_law_1.2,0.08476160049438478
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,1,64,power_law_1.2,0.13852288246154784
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,1,64,power_law_1.2,0.1385536003112793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,1,64,power_law_1.2,0.1492569637298584
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,1,64,power_law_1.2,0.14895872116088865
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,1,64,power_law_1.01,0.9309081268310546
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,1,64,power_law_1.2,0.15010047912597657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,1,64,power_law_1.2,0.1512204837799072
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,1,64,power_law_1.2,0.15539584159851075
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,1,64,power_law_1.2,0.15917056083679199
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,1,64,power_law_1.2,0.16424959182739257
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,1,64,power_law_1.01,1.1712345886230469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,1,64,power_law_1.2,0.17378944396972656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,1,64,power_law_1.2,0.17743999481201173
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,1,64,power_law_1.2,0.1782156753540039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,1,64,power_law_1.2,0.19576704025268554
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,1,64,power_law_1.2,0.24849664688110354
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,1,64,power_law_1.2,0.2884236717224121
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,1,64,power_law_1.2,0.3626496124267578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,1,64,power_law_1.2,0.5566515350341797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,1,64,power_law_1.2,0.6697958374023438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,1,64,power_law_1.2,0.06066559791564942
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,1,64,power_law_1.2,0.059905281066894536
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,1,64,power_law_1.2,0.973043212890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,1,64,power_law_1.2,0.061053438186645506
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,1,64,power_law_1.2,0.062280960083007805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,1,64,power_law_1.2,0.06492671966552735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,1,64,power_law_1.2,1.341370849609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,1,64,power_law_1.2,0.06850944042205812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,1,64,power_law_1.2,0.06944767951965332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,1,64,power_law_1.2,0.07322624206542969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,1,64,power_law_1.2,0.07236735820770264
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,1,64,power_law_1.2,1.8333529663085937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,1,64,power_law_1.2,0.07860991954803467
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,1,64,power_law_1.2,0.08033920288085937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,1,64,power_law_1.2,0.08964735984802245
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,1,64,power_law_1.2,0.06050559997558593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,1,64,power_law_1.2,0.09144960403442383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,1,64,power_law_1.2,0.09761024475097656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,1,64,power_law_1.2,0.11619711875915528
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,1,64,power_law_1.2,0.13584768295288085
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,1,64,power_law_1.2,0.06735104084014892
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,1,64,power_law_1.2,0.17849599838256835
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,1,64,power_law_1.2,0.24252672195434571
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,1,64,power_law_1.2,0.31818624496459963
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,1,64,power_law_1.2,0.38947200775146484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,2,32,balanced,0.042940158843994145
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,1,64,power_law_1.2,0.666192626953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,1,64,power_law_1.2,0.08756992340087891
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,2,32,balanced,0.042780160903930664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,1,64,power_law_1.2,3.522746887207031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,2,32,balanced,0.04454271793365479
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,2,32,balanced,0.056723198890686034
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,1,64,power_law_1.2,0.9539673614501953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,2,32,balanced,0.06757247924804688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,2,32,balanced,0.10079487800598144
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,2,32,balanced,0.14200575828552248
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,2,32,balanced,0.14348928451538084
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,2,32,balanced,0.14967424392700196
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,1,64,power_law_1.2,1.3706393432617188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,2,32,balanced,0.15330304145812987
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,2,32,balanced,0.15884415626525877
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,1,64,power_law_1.2,4.54951171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,2,32,balanced,0.16202239990234374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,2,32,balanced,0.15353983879089356
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,1,64,power_law_1.2,1.6744358825683594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,2,32,balanced,0.17171712875366213
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,2,32,balanced,0.16462335586547852
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,2,32,balanced,0.19074176788330077
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,2,32,balanced,0.1993177604675293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,2,32,balanced,0.1438361644744873
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,2,32,balanced,0.1449766445159912
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,2,32,balanced,0.2741004753112793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,2,32,balanced,0.144780797958374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,2,32,balanced,0.1512947177886963
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,2,32,balanced,0.4087628936767578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,2,32,balanced,0.030280959606170655
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,2,32,balanced,0.4823731231689453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,2,32,balanced,0.03020416021347046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,2,32,balanced,0.03485440015792847
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,2,32,balanced,0.04473855972290039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,2,32,balanced,0.06288127899169922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,2,32,balanced,0.6853427124023438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,2,32,balanced,0.07116415977478027
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,2,32,balanced,0.07326720237731935
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,2,32,balanced,0.07261312007904053
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,2,32,balanced,0.8789965057373047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,2,32,balanced,0.07353983879089356
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,2,32,balanced,0.07378176212310791
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,2,32,balanced,0.07465983867645264
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,2,32,balanced,0.07618303775787354
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,2,32,balanced,0.25299968719482424
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,2,32,balanced,0.08240896224975586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,2,32,balanced,0.08497535705566406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,2,32,balanced,0.07847040176391602
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,2,32,balanced,0.08551936149597168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,2,32,balanced,0.09209983825683594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,2,32,balanced,0.10263808250427246
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,2,32,balanced,0.11638400077819824
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,2,32,balanced,0.13995391845703126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,2,32,balanced,0.029879040718078613
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,2,32,balanced,0.21795839309692383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,2,32,balanced,0.2636735916137695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,2,32,power_law_1.01,0.09385343551635741
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,2,32,balanced,0.38273536682128906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,2,32,power_law_1.01,0.0933836841583252
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,2,32,balanced,0.4750835037231445
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,2,32,power_law_1.01,0.09412991523742675
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,2,32,balanced,0.07852287769317627
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,2,32,power_law_1.01,0.1316531181335449
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,2,32,power_law_1.01,0.13691519737243651
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,2,32,power_law_1.01,0.14074496269226072
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,2,32,power_law_1.01,0.1471782398223877
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,2,32,balanced,0.16301952362060548
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,2,32,power_law_1.01,0.15106944084167478
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,2,32,power_law_1.01,0.15294719696044923
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,2,32,power_law_1.01,0.15857151985168455
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,2,32,power_law_1.01,0.15454079627990722
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,2,32,power_law_1.01,0.15608960151672363
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,2,32,power_law_1.01,0.09347455978393555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,2,32,power_law_1.01,0.19949695587158203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,2,32,power_law_1.01,0.22758655548095702
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,2,32,power_law_1.01,0.12951040267944336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,2,32,power_law_1.01,0.27238271713256834
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,2,32,power_law_1.01,0.36807167053222656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,2,32,power_law_1.01,0.135600643157959
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,2,32,power_law_1.01,0.4695334243774414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,2,32,power_law_1.01,0.6056934356689453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,2,32,power_law_1.01,0.15757056236267092
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,2,32,power_law_1.01,0.059613437652587885
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,2,32,power_law_1.01,0.7831423950195313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,2,32,power_law_1.01,0.16985599517822264
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,2,32,power_law_1.01,0.06123519897460937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,2,32,power_law_1.01,0.06207744121551514
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,2,32,power_law_1.01,0.06242559909820557
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,2,32,power_law_1.01,1.0874188995361327
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,2,32,power_law_1.01,0.06722047805786133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,2,32,power_law_1.01,0.06961408138275146
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,2,32,power_law_1.01,0.07498112201690674
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,2,32,power_law_1.01,0.07126527786254883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,2,32,power_law_1.01,0.07421696186065674
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,2,32,power_law_1.01,0.07642111778259278
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,2,32,power_law_1.01,0.07789951801300049
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,2,32,power_law_1.01,0.08035584449768066
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,2,32,power_law_1.01,0.08137984275817871
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,2,32,power_law_1.01,0.08636287689208985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,2,32,power_law_1.01,0.09898367881774903
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,2,32,power_law_1.01,1.7113958740234374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,2,32,power_law_1.01,0.11132927894592286
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,2,32,power_law_1.01,0.15305088043212892
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,2,32,power_law_1.01,0.1785087966918945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,2,32,power_law_1.01,2.0626048278808593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,2,32,power_law_1.01,0.26030975341796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,2,32,power_law_1.01,0.3163737678527832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,2,32,power_law_1.01,0.4458099365234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,2,32,power_law_1.2,0.09338496208190919
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,2,32,power_law_1.01,0.08256511688232422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,2,32,power_law_1.01,0.6091686248779297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,2,32,power_law_1.2,0.09293055534362793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,2,32,power_law_1.2,0.09374079704284669
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,2,32,power_law_1.2,0.09370112419128418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,2,32,power_law_1.2,0.1262476825714111
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,2,32,power_law_1.01,0.8874905395507813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,2,32,power_law_1.2,0.12874367713928223
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,2,32,power_law_1.01,0.06601984024047851
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,2,32,power_law_1.01,1.0957810974121094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,2,32,power_law_1.2,0.14520319938659668
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,2,32,power_law_1.2,0.13914496421813966
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,2,32,power_law_1.2,0.14575615882873535
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,2,32,power_law_1.2,0.1451097583770752
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,2,32,power_law_1.2,0.14968832015991212
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,2,32,power_law_1.2,0.1553664016723633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,2,32,power_law_1.01,0.0820736026763916
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,2,32,power_law_1.2,0.153818883895874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,2,32,power_law_1.2,0.17837440490722656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,2,32,power_law_1.2,0.221661434173584
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,2,32,power_law_1.2,0.24879487991333007
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,2,32,power_law_1.2,0.31589632034301757
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,2,32,power_law_1.2,0.14117888450622557
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,2,32,power_law_1.2,0.1417088031768799
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,2,32,power_law_1.2,0.4010457611083984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,2,32,power_law_1.2,0.5384665679931641
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,2,32,power_law_1.2,0.7137203216552734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,2,32,power_law_1.2,0.1571072006225586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,2,32,power_law_1.2,0.05947391986846924
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,2,32,power_law_1.2,0.061326079368591316
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,2,32,power_law_1.2,0.061306881904602054
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,2,32,power_law_1.2,0.062449917793273926
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,2,32,power_law_1.2,1.0802982330322266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,2,32,power_law_1.2,0.06333568096160888
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,2,32,power_law_1.2,0.06377471923828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,2,32,power_law_1.2,0.07331840038299561
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,2,32,power_law_1.2,0.07253759860992431
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,2,32,power_law_1.2,0.07159808158874512
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,2,32,power_law_1.2,1.5413095092773437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,2,32,power_law_1.2,0.07713280200958253
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,2,32,power_law_1.2,0.08012160301208496
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,2,32,power_law_1.2,0.08284416198730468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,2,32,power_law_1.2,0.08082304000854493
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,2,32,power_law_1.2,0.08505727767944335
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,2,32,power_law_1.2,0.09304832458496094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,2,32,power_law_1.2,0.10883328437805176
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,2,32,power_law_1.2,0.11947648048400879
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,2,32,power_law_1.2,0.174597110748291
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,2,32,power_law_1.2,0.22407039642333984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,2,32,power_law_1.2,0.2881843185424805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,2,32,power_law_1.2,0.07219071865081787
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,2,32,power_law_1.2,0.39348224639892576
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,2,32,power_law_1.2,0.5555788803100585
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,4,16,balanced,0.02861696004867554
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,4,16,balanced,0.02888319969177246
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,4,16,balanced,0.04809216022491455
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,2,32,power_law_1.2,3.309879150390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,4,16,balanced,0.05667840003967285
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,4,16,balanced,0.06776063919067384
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,2,32,power_law_1.2,0.802059555053711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,4,16,balanced,0.1044377613067627
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,4,16,balanced,0.14687104225158693
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,2,32,power_law_1.2,2.5329356384277344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,4,16,balanced,0.15172096252441408
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,2,32,power_law_1.2,1.0572185516357422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,2,32,power_law_1.2,0.07601408004760743
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,4,16,balanced,0.16108671188354493
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,4,16,balanced,0.15220095634460448
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,4,16,balanced,0.15798272132873534
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,4,16,balanced,0.15970175743103027
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,4,16,balanced,0.16557056427001954
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,2,32,power_law_1.2,1.5326220703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,4,16,balanced,0.1769843292236328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,4,16,balanced,0.18860544204711913
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,4,16,balanced,0.21337856292724608
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,4,16,balanced,0.2306777572631836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,4,16,balanced,0.2956211280822754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,4,16,balanced,0.33518463134765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,4,16,balanced,0.1468057632446289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,4,16,balanced,0.15248255729675292
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,4,16,balanced,0.1533670425415039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,4,16,balanced,0.029192960262298583
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,4,16,balanced,0.5104127883911133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,4,16,balanced,0.1574630355834961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,4,16,balanced,0.02920959949493408
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,4,16,balanced,0.02962048053741455
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,4,16,balanced,0.047861762046813965
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,4,16,balanced,0.06662144184112548
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,4,16,balanced,0.08001664161682129
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,4,16,balanced,0.07750783920288086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,4,16,balanced,0.07964159965515136
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,4,16,balanced,0.08383872032165526
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,4,16,balanced,0.08574720382690429
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,4,16,balanced,0.0886195182800293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,4,16,balanced,0.08224512100219726
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,4,16,balanced,0.08446463584899902
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,4,16,balanced,1.1571238708496092
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,4,16,balanced,0.09262335777282715
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,4,16,balanced,0.10313728332519531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,4,16,balanced,0.11518079757690429
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,4,16,balanced,0.1357107162475586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,4,16,balanced,0.1558182430267334
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,4,16,balanced,0.6241049575805664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,4,16,balanced,0.1953984069824219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,4,16,balanced,0.23378047943115235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,4,16,balanced,0.07988096237182617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,4,16,balanced,0.8892403411865235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,4,16,balanced,0.32844799041748046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,4,16,balanced,0.4063782501220703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,4,16,power_law_1.01,0.06662784099578858
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,4,16,power_law_1.01,0.09380607604980469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,4,16,balanced,0.08702207565307617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,4,16,power_law_1.01,0.09232383728027344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,4,16,balanced,0.6015206527709961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,4,16,power_law_1.01,0.09488896369934083
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,4,16,power_law_1.01,0.09398528099060058
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,4,16,balanced,0.7651583862304687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,4,16,power_law_1.01,0.1303769588470459
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,4,16,power_law_1.01,0.13651328086853026
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,4,16,power_law_1.01,0.12846976280212402
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,4,16,power_law_1.01,0.14154751777648927
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,4,16,balanced,0.03461888074874878
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,4,16,power_law_1.01,0.1461043167114258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,4,16,power_law_1.01,0.14378623962402343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,4,16,power_law_1.01,0.15476351737976074
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,4,16,power_law_1.01,0.15569791793823243
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,4,16,power_law_1.01,0.14997759819030762
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,4,16,power_law_1.01,0.15955967903137208
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,4,16,power_law_1.01,0.16346752166748046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,4,16,power_law_1.01,0.18663423538208007
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,4,16,power_law_1.01,0.20543487548828124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,4,16,power_law_1.01,0.276364803314209
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,4,16,power_law_1.01,0.3320127868652344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,4,16,power_law_1.01,0.4292671966552734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,4,16,power_law_1.01,0.5511667251586914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,4,16,power_law_1.01,0.04648575782775879
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,4,16,power_law_1.01,0.06234879970550537
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,4,16,power_law_1.01,0.15714303970336913
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,4,16,power_law_1.01,0.061665282249450684
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,4,16,power_law_1.01,1.044000015258789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,4,16,power_law_1.01,0.06361599922180175
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,4,16,power_law_1.01,0.06699007987976074
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,4,16,power_law_1.01,0.07686783790588378
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,4,16,power_law_1.01,0.07247615814208984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,4,16,power_law_1.01,0.07424895763397217
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,4,16,power_law_1.01,0.07494016170501709
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,4,16,power_law_1.01,0.08163455963134766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,4,16,power_law_1.01,0.8056588745117187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,4,16,power_law_1.01,1.8402919006347656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,4,16,power_law_1.01,0.08215295791625976
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,4,16,power_law_1.01,0.0789299201965332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,4,16,power_law_1.01,0.06325119972229004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,4,16,power_law_1.01,0.08198143959045409
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,4,16,power_law_1.01,0.08665727615356446
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,4,16,power_law_1.01,0.11217791557312011
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,4,16,power_law_1.01,0.12882304191589355
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,4,16,power_law_1.01,0.16035455703735352
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,4,16,power_law_1.01,1.3817855834960937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,4,16,power_law_1.01,0.19772415161132811
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,4,16,power_law_1.01,0.2773772811889649
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,4,16,power_law_1.01,0.07916800022125245
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,4,16,power_law_1.01,0.3632179260253906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,4,16,power_law_1.2,0.06707200050354004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,4,16,power_law_1.01,0.5143743896484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,4,16,power_law_1.2,0.09214847564697266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,4,16,power_law_1.2,0.08661248207092284
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,4,16,power_law_1.01,0.6269158554077149
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,4,16,power_law_1.2,0.09145088195800781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,4,16,power_law_1.01,0.09422719955444336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,4,16,power_law_1.2,0.09521280288696289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,4,16,power_law_1.2,0.1266380786895752
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,4,16,power_law_1.2,0.13506688117980956
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,4,16,power_law_1.01,0.8581721496582031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,4,16,power_law_1.2,0.13233920097351076
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,4,16,power_law_1.2,0.14258560180664062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,4,16,power_law_1.2,0.13938176155090332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,4,16,power_law_1.2,0.151147518157959
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,4,16,power_law_1.2,0.15309823989868163
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,4,16,power_law_1.2,0.15336447715759277
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,4,16,power_law_1.01,1.139920654296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,4,16,power_law_1.2,0.1574284839630127
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,4,16,power_law_1.2,0.14690431594848632
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,4,16,power_law_1.2,0.1717862319946289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,4,16,power_law_1.2,0.19650815963745116
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,4,16,power_law_1.2,0.22688255310058594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,4,16,power_law_1.2,0.2778867149353027
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,4,16,power_law_1.2,0.3737625503540039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,4,16,power_law_1.2,0.4845171356201172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,4,16,power_law_1.2,0.7024345397949219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,4,16,power_law_1.2,0.04696320056915283
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,4,16,power_law_1.2,0.062233600616455075
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,4,16,power_law_1.2,0.060002560615539546
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,4,16,power_law_1.2,0.06374400138854981
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,4,16,power_law_1.2,0.06484096050262451
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,4,16,power_law_1.2,0.9725350189208986
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,4,16,power_law_1.2,0.0644979190826416
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,4,16,power_law_1.2,0.07110400199890136
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,4,16,power_law_1.2,0.07197311878204346
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,4,16,power_law_1.2,0.07630720138549804
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,4,16,power_law_1.2,0.07416063785552979
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,4,16,power_law_1.2,0.07821055889129638
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,4,16,power_law_1.2,0.08077183723449707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,4,16,power_law_1.2,0.08559871673583984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,4,16,power_law_1.2,0.07911295890808105
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,4,16,power_law_1.2,1.8055052185058593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,4,16,power_law_1.2,0.08920448303222657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,4,16,power_law_1.2,0.09441791534423828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,4,16,power_law_1.2,0.16077568054199218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,4,16,power_law_1.2,0.12005632400512696
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,4,16,power_law_1.2,0.13604479789733886
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,4,16,power_law_1.2,0.17424127578735352
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,4,16,power_law_1.2,0.21992191314697265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,4,16,power_law_1.2,1.2565811157226563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,4,16,power_law_1.2,2.4861952209472653
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,4,16,power_law_1.2,0.29583744049072264
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,4,16,power_law_1.2,0.40218624114990237
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,8,8,balanced,0.028510720729827882
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,8,8,balanced,0.029672958850860597
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,8,8,balanced,0.04944896221160889
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,4,16,power_law_1.2,0.08620544433593749
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,4,16,power_law_1.2,0.7405094146728516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,8,8,balanced,0.07185535907745362
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,8,8,balanced,0.10375295639038087
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,8,8,balanced,0.1485247993469238
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,8,8,balanced,0.15398271560668947
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,8,8,balanced,0.1567859172821045
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,8,8,balanced,0.14919039726257324
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,8,8,balanced,0.15306367874145507
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,8,8,balanced,0.15586048126220703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,4,16,power_law_1.2,1.5021133422851562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,8,8,balanced,0.16158464431762695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,4,16,power_law_1.2,0.5939660644531249
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,8,8,balanced,0.16611328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,8,8,balanced,0.05816319942474365
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,8,8,balanced,0.18001407623291016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,8,8,balanced,0.19833087921142578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,8,8,balanced,0.21681152343750001
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,8,8,balanced,0.15208319664001463
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,8,8,balanced,0.25649280548095704
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,4,16,power_law_1.2,1.057290267944336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,8,8,balanced,0.29513856887817386
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,8,8,balanced,0.40097408294677733
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,8,8,balanced,0.4771046447753906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,8,8,balanced,0.02916863918304443
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,8,8,balanced,0.029780480861663815
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,8,8,balanced,0.03201280117034912
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,8,8,balanced,0.03966336011886597
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,8,8,balanced,0.0528000020980835
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,8,8,balanced,0.17203968048095702
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,8,8,balanced,0.09171711921691894
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,8,8,balanced,0.9174771118164063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,8,8,balanced,0.0945958423614502
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,8,8,balanced,0.09790719985961914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,8,8,balanced,0.09980031967163086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,8,8,balanced,0.09201408386230468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,8,8,balanced,0.09533056259155273
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,8,8,balanced,0.09732735633850098
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,8,8,balanced,0.10126463890075683
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,8,8,balanced,1.308343048095703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,8,8,balanced,0.10477439880371094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,8,8,balanced,0.11052672386169433
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,8,8,balanced,0.12023167610168457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,8,8,balanced,0.14109951972961426
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,8,8,balanced,0.15931648254394531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,8,8,balanced,0.19525503158569335
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,8,8,balanced,1.692560577392578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,8,8,balanced,0.2330342483520508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,8,8,balanced,0.7268377685546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,8,8,balanced,0.3069580841064453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,8,8,balanced,0.380081901550293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,8,8,balanced,0.07666687965393067
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,8,8,power_law_1.01,0.056734719276428223
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,8,8,power_law_1.01,0.06883967876434326
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,8,8,balanced,0.5534668731689453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,8,8,power_law_1.01,0.07314176082611083
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,8,8,power_law_1.01,0.07970687866210938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,8,8,balanced,0.6996838378906249
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,8,8,power_law_1.01,0.0865664005279541
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,8,8,power_law_1.01,0.08546560287475587
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,8,8,power_law_1.01,0.13301888465881348
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,8,8,power_law_1.01,0.1357094383239746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,8,8,power_law_1.01,0.13785216331481934
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,8,8,power_law_1.01,0.14799103736877442
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,8,8,balanced,1.038071060180664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,8,8,power_law_1.01,0.14705535888671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,8,8,power_law_1.01,0.1437478446960449
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,8,8,power_law_1.01,0.14818304061889648
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,8,8,power_law_1.01,0.15394432067871094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,8,8,power_law_1.01,0.1583129596710205
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,8,8,balanced,1.3507431030273438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,8,8,power_law_1.01,0.16515840530395506
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,8,8,power_law_1.01,0.18945280075073243
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,8,8,power_law_1.01,0.2048793601989746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,8,8,power_law_1.01,0.24551040649414063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,8,8,power_law_1.01,0.3059174346923828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,8,8,power_law_1.01,0.4219507217407227
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,8,8,power_law_1.01,0.037730560302734376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,8,8,power_law_1.01,0.5285823822021485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,8,8,power_law_1.01,0.049396481513977054
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,8,8,power_law_1.01,0.7445849609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,8,8,power_law_1.01,0.053442559242248534
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,8,8,power_law_1.01,0.05910783767700195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,8,8,power_law_1.01,0.0631769609451294
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,8,8,power_law_1.01,0.06258304119110107
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,8,8,power_law_1.01,0.08218239784240723
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,8,8,power_law_1.01,0.08572799682617187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,8,8,power_law_1.01,0.08791040420532227
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,8,8,power_law_1.01,0.08834048271179198
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,8,8,power_law_1.01,0.09331071853637696
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,8,8,power_law_1.01,0.09385727882385254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,8,8,power_law_1.01,0.08941439628601075
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,8,8,power_law_1.01,0.09454976081848145
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,8,8,power_law_1.01,0.09829248428344727
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,8,8,power_law_1.01,0.10021247863769531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,8,8,power_law_1.01,0.14213760375976564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,8,8,power_law_1.01,1.7993075561523437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,8,8,power_law_1.01,0.1294668769836426
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,8,8,power_law_1.01,0.14722687721252442
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,8,8,power_law_1.01,0.9382335662841796
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,8,8,power_law_1.01,0.18033151626586913
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,8,8,power_law_1.01,0.21882240295410157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,8,8,power_law_1.01,0.27905664443969724
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,8,8,power_law_1.01,0.3797119903564453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,8,8,power_law_1.01,1.5319999694824218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,8,8,power_law_1.2,0.05681920051574707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,8,8,power_law_1.01,0.6860364532470703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,8,8,power_law_1.2,0.07362175941467285
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,8,8,power_law_1.2,0.06922368049621583
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,8,8,power_law_1.2,0.07817215919494629
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,8,8,power_law_1.2,0.0823078441619873
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,8,8,power_law_1.2,0.09095552444458008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,8,8,power_law_1.2,0.13414015769958496
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,8,8,power_law_1.01,1.275331802368164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,8,8,power_law_1.2,0.13907584190368652
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,8,8,power_law_1.01,0.5281395339965821
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,8,8,power_law_1.2,0.1435097599029541
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,8,8,power_law_1.2,0.14547455787658692
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,8,8,power_law_1.2,0.14936448097229005
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,8,8,power_law_1.01,0.10641280174255371
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,8,8,power_law_1.2,0.1418227195739746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,8,8,power_law_1.2,0.15041152000427246
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,8,8,power_law_1.2,0.156495361328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,8,8,power_law_1.2,0.15963647842407228
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,8,8,power_law_1.2,0.17005823135375978
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,8,8,power_law_1.2,0.19377151489257813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,8,8,power_law_1.2,0.21750015258789063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,8,8,power_law_1.2,0.2717862319946289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,8,8,power_law_1.2,0.3179596710205078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,8,8,power_law_1.2,0.13714943885803224
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,8,8,power_law_1.2,0.458100471496582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,8,8,power_law_1.2,0.03764607906341553
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,8,8,power_law_1.2,0.6054515075683594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,8,8,power_law_1.2,0.04928768157958984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,8,8,power_law_1.2,0.049077758789062495
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,8,8,power_law_1.2,0.8123020935058595
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,8,8,power_law_1.01,0.9721804809570311
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,8,8,power_law_1.2,0.06264832019805908
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,8,8,power_law_1.2,1.0667430114746093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,8,8,power_law_1.2,0.06702720165252686
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,8,8,power_law_1.2,0.08329471588134765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,8,8,power_law_1.2,0.08772735595703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,8,8,power_law_1.2,0.08824064254760741
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,8,8,power_law_1.2,0.08939519882202149
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,8,8,power_law_1.2,0.094967041015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,8,8,power_law_1.2,0.09408767700195313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,8,8,power_law_1.2,0.08866175651550293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,8,8,power_law_1.2,0.09507712364196777
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,8,8,power_law_1.2,0.09805952072143555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,8,8,power_law_1.2,1.5776806640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,8,8,power_law_1.2,0.10129535675048829
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,8,8,power_law_1.2,0.10989312171936035
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,8,8,power_law_1.2,0.13398271560668945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,8,8,power_law_1.2,0.14508543968200685
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,8,8,power_law_1.2,0.05555456161499024
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,8,8,power_law_1.2,0.22234880447387698
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,8,8,power_law_1.2,0.31288320541381837
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,8,8,power_law_1.2,2.0684915161132813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,8,8,power_law_1.2,0.41613311767578126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,16,4,balanced,0.027740159034729005
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,16,4,balanced,0.030538239479064942
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,16,4,balanced,0.06269440174102783
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,8,8,power_law_1.2,0.5654156875610352
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,16,4,balanced,0.10877951622009277
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,8,8,power_law_1.2,0.7328038024902344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,16,4,balanced,0.10431360244750976
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,16,4,balanced,0.1070246410369873
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,16,4,balanced,0.14792192459106446
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,8,8,power_law_1.2,0.18619007110595703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,16,4,balanced,0.15060992240905763
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,16,4,balanced,0.15373567581176756
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,16,4,balanced,0.15561471939086916
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,16,4,balanced,0.16139263153076172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,16,4,balanced,0.16407167434692382
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,8,8,power_law_1.2,1.4378790283203124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,16,4,balanced,0.17013631820678712
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,16,4,balanced,0.1787046432495117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,16,4,balanced,0.05078015804290772
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,16,4,balanced,0.19609216690063475
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,16,4,balanced,0.2494451141357422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,16,4,balanced,0.2878144073486328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,16,4,balanced,0.3837555313110351
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,16,4,balanced,0.1473523235321045
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,16,4,balanced,0.463768310546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,8,8,power_law_1.2,1.090655975341797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,16,4,balanced,0.028337919712066652
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,16,4,balanced,0.7126284790039062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,16,4,balanced,0.03312511920928955
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,16,4,balanced,0.029881598949432375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,16,4,balanced,0.07410304069519043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,16,4,balanced,0.05949056148529053
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,16,4,balanced,0.08751359939575196
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,16,4,balanced,0.9454118347167968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,16,4,balanced,0.2126515197753906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,16,4,balanced,0.08864895820617676
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,16,4,balanced,0.09026432037353516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,16,4,balanced,0.11342464447021486
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,16,4,balanced,0.11355263710021972
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,16,4,balanced,0.11625727653503418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,16,4,balanced,0.1187673568725586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,16,4,balanced,0.12070272445678712
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,16,4,balanced,0.12522368431091307
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,16,4,balanced,1.337592315673828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,16,4,balanced,0.12729087829589844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,16,4,balanced,0.13311360359191896
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,16,4,balanced,0.14034687995910644
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,16,4,balanced,0.15951744079589844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,16,4,balanced,0.1754457664489746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,16,4,balanced,0.21324928283691405
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,16,4,balanced,1.7657075500488282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,16,4,balanced,0.24639104843139648
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,16,4,balanced,0.31590784072875977
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,16,4,balanced,0.043807997703552246
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,16,4,balanced,0.3864255905151367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,16,4,power_law_1.01,0.03071232080459595
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,16,4,power_law_1.01,0.05749375820159912
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,16,4,balanced,0.5746726226806641
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,16,4,power_law_1.01,0.06198400020599365
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,16,4,power_law_1.01,0.07018367767333984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,16,4,power_law_1.01,0.07691648006439208
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,16,4,power_law_1.01,0.08265472412109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,16,4,power_law_1.01,0.1016716766357422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,16,4,power_law_1.01,0.09705727577209473
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,16,4,power_law_1.01,0.129932804107666
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,16,4,balanced,1.0675955200195313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,16,4,power_law_1.01,0.1378547191619873
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,16,4,power_law_1.01,0.14135295867919923
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,16,4,power_law_1.01,0.14517120361328123
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,16,4,balanced,1.3941491699218749
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,16,4,power_law_1.01,0.14916607856750488
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,16,4,power_law_1.01,0.15204480171203613
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,16,4,power_law_1.01,0.15999487876892088
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,16,4,power_law_1.01,0.17035903930664062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,16,4,power_law_1.01,0.2104857635498047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,16,4,balanced,0.7163827514648438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,16,4,power_law_1.01,0.2524083137512207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,16,4,power_law_1.01,0.29563135147094727
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,16,4,power_law_1.01,0.400945930480957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,16,4,power_law_1.01,0.13326080322265627
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,16,4,power_law_1.01,0.5224396896362304
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,16,4,power_law_1.01,0.0325708794593811
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,16,4,power_law_1.01,0.039696640968322754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,16,4,power_law_1.01,0.04945151805877686
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,16,4,power_law_1.01,0.7781017303466797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,16,4,power_law_1.01,0.056339201927185056
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,16,4,power_law_1.01,0.9739071655273438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,16,4,power_law_1.01,0.06121471881866455
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,16,4,power_law_1.01,0.06594560146331788
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,16,4,power_law_1.01,0.08535679817199707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,16,4,power_law_1.01,0.0822156810760498
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,16,4,power_law_1.01,0.09577343940734864
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,16,4,power_law_1.01,0.09787520408630371
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,16,4,power_law_1.01,0.10161919593811035
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,16,4,power_law_1.01,0.10477567672729493
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,16,4,power_law_1.01,1.3918911743164064
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,16,4,power_law_1.01,0.11433600425720214
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,16,4,power_law_1.01,0.11875328063964843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,16,4,power_law_1.01,0.12338560104370118
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,16,4,power_law_1.01,0.13133695602416992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,16,4,power_law_1.01,0.15331839561462401
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,16,4,power_law_1.01,0.17337087631225584
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,16,4,power_law_1.01,0.21587455749511716
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,16,4,power_law_1.01,0.26291072845458985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,16,4,power_law_1.01,1.8387161254882813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,16,4,power_law_1.01,0.1939583969116211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,16,4,power_law_1.01,0.3432460784912109
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,16,4,power_law_1.2,0.03048959970474243
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,16,4,power_law_1.01,0.443559684753418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,16,4,power_law_1.2,0.05680511951446533
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,16,4,power_law_1.01,0.6186495971679687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,16,4,power_law_1.2,0.05992063999176025
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,16,4,power_law_1.01,0.11134719848632812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,16,4,power_law_1.2,0.06896512031555176
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,16,4,power_law_1.01,0.7700032043457031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,16,4,power_law_1.2,0.07523327827453613
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,16,4,power_law_1.2,0.09701888084411621
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,16,4,power_law_1.2,0.08253312110900879
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,16,4,power_law_1.2,0.13084287643432618
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,16,4,power_law_1.2,0.13203712463378908
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,16,4,power_law_1.2,0.14206720352172852
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,16,4,power_law_1.2,0.149486083984375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,16,4,power_law_1.2,0.15550080299377442
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,16,4,power_law_1.2,0.16203519821166992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,16,4,power_law_1.01,1.4771839904785158
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,16,4,power_law_1.2,0.17405311584472655
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,16,4,power_law_1.2,0.19496063232421873
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,16,4,power_law_1.2,0.2157119941711426
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,16,4,power_law_1.2,0.3121331214904785
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,16,4,power_law_1.2,0.09287808418273927
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,16,4,power_law_1.2,0.13784064292907713
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,16,4,power_law_1.01,1.1256063842773438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,16,4,power_law_1.2,0.5667084884643555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,16,4,power_law_1.2,0.032627201080322264
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,16,4,power_law_1.2,0.7843865966796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,16,4,power_law_1.2,0.04037119865417481
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,16,4,power_law_1.2,0.2630233573913574
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,16,4,power_law_1.2,1.0450879669189452
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,16,4,power_law_1.2,0.045239038467407226
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,16,4,power_law_1.2,0.42382591247558593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,16,4,power_law_1.2,0.05379712104797363
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,16,4,power_law_1.2,0.06163712024688721
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,16,4,power_law_1.2,0.0622489595413208
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,16,4,power_law_1.2,0.0807436752319336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,16,4,power_law_1.2,0.07765247821807861
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,16,4,power_law_1.2,0.09544192314147949
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,16,4,power_law_1.2,0.0981158447265625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,16,4,power_law_1.2,0.1011520004272461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,16,4,power_law_1.2,0.10527872085571288
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,16,4,power_law_1.2,0.10926079750061035
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,16,4,power_law_1.2,0.1194803237915039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,16,4,power_law_1.2,0.1251302433013916
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,16,4,power_law_1.2,0.13678848266601562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,16,4,power_law_1.2,0.13960448265075684
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,16,4,power_law_1.2,0.159749116897583
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,16,4,power_law_1.2,0.17856767654418945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,16,4,power_law_1.2,0.22340351104736328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,16,4,power_law_1.2,1.4692454528808594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,16,4,power_law_1.2,0.26675455093383793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,16,4,power_law_1.2,0.34375553131103515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,16,4,power_law_1.2,0.43564414978027344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,32,2,balanced,0.02870784044265747
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,16,4,power_law_1.2,2.0539263916015624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,16,4,power_law_1.2,0.6436544036865234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,32,2,balanced,0.030307838916778563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,32,2,balanced,0.05223552227020264
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,32,2,balanced,0.06441215991973877
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,16,4,power_law_1.2,0.8201382446289063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,32,2,balanced,0.08603903770446777
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,32,2,balanced,0.12144895553588868
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,32,2,balanced,0.12587648391723633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,32,2,balanced,0.12384639739990236
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,32,2,balanced,0.12889087677001954
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,16,4,power_law_1.2,1.1459455871582032
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,32,2,balanced,0.16378368377685545
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,32,2,balanced,0.1656384086608887
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,16,4,power_law_1.2,0.11619071960449219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,32,2,balanced,0.17190912246704101
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,32,2,balanced,0.17748096466064453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,32,2,balanced,0.1840025520324707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,32,2,balanced,0.20274431228637696
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,32,2,balanced,0.21859840393066404
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,16,4,power_law_1.2,1.574694366455078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,32,2,balanced,0.25466495513916015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,32,2,balanced,0.2916326332092285
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,32,2,balanced,0.3858892822265625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,32,2,balanced,0.12305536270141601
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,32,2,balanced,0.12482303619384764
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,32,2,balanced,0.45999614715576176
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,32,2,balanced,0.029390079975128176
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,32,2,balanced,0.032497920989990235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,32,2,balanced,0.037607679367065425
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,32,2,balanced,0.7095423889160156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,32,2,balanced,0.1685478401184082
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,32,2,balanced,0.05308671951293945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,32,2,balanced,0.08332032203674317
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,32,2,balanced,0.12474111557006835
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,32,2,balanced,0.1256332778930664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,32,2,balanced,0.9443532562255859
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,32,2,balanced,0.12925824165344238
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,32,2,balanced,0.13049087524414063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,32,2,balanced,0.1324351978302002
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,32,2,balanced,0.17308160781860352
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,32,2,balanced,0.17904767990112305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,32,2,balanced,0.17547903060913086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,32,2,balanced,0.18159616470336915
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,32,2,balanced,0.18640640258789062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,32,2,balanced,0.19221504211425783
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,32,2,balanced,0.20913663864135743
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,32,2,balanced,0.2212428855895996
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,32,2,balanced,0.25601791381835937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,32,2,balanced,0.2874176025390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,32,2,balanced,0.3526873779296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,32,2,balanced,0.4194291305541992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,32,2,balanced,1.8875442504882813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,32,2,balanced,0.12778112411499024
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,32,2,balanced,0.6721011352539062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,32,2,power_law_1.01,0.02997119903564453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,32,2,power_law_1.01,0.05043456077575683
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,32,2,balanced,0.8076428985595703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,32,2,power_law_1.01,0.06706560134887696
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,32,2,balanced,1.4071615600585938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,32,2,power_law_1.01,0.09223423957824707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,32,2,power_law_1.01,0.10383487701416017
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,32,2,power_law_1.01,0.10811264038085937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,32,2,power_law_1.01,0.11035136222839355
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,32,2,balanced,1.1984947204589844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,32,2,power_law_1.01,0.11426688194274903
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,32,2,power_law_1.01,0.12173952102661131
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,32,2,power_law_1.01,0.15210623741149903
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,32,2,power_law_1.01,0.15509887695312502
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,32,2,power_law_1.01,0.1620364761352539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,32,2,balanced,1.586588134765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,32,2,power_law_1.01,0.1778278350830078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,32,2,power_law_1.01,0.05672448158264161
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,32,2,power_law_1.01,0.20227840423583981
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,32,2,power_law_1.01,0.2205824089050293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,32,2,power_law_1.01,0.08434687614440918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,32,2,power_law_1.01,0.2651136016845703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,32,2,power_law_1.01,0.3057779121398926
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,32,2,power_law_1.01,0.3911206436157227
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,32,2,power_law_1.01,0.5138585662841797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,32,2,power_law_1.01,0.15850111961364746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,32,2,power_law_1.01,0.03192960023880005
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,32,2,power_law_1.01,0.7365811157226563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,32,2,power_law_1.01,0.03633280038833618
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,32,2,power_law_1.01,0.045781760215759276
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,32,2,power_law_1.01,0.9786726379394531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,32,2,power_law_1.01,0.06062592029571533
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,32,2,power_law_1.01,0.07012735843658448
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,32,2,power_law_1.01,0.0821350383758545
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,32,2,power_law_1.01,0.1031935977935791
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,32,2,power_law_1.01,1.3856524658203124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,32,2,power_law_1.01,0.10636672019958496
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,32,2,power_law_1.01,0.10926591873168945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,32,2,power_law_1.01,0.11299327850341796
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,32,2,power_law_1.01,0.1554035186767578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,32,2,power_law_1.01,0.15872384071350099
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,32,2,power_law_1.01,0.16980735778808592
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,32,2,power_law_1.01,0.16219263076782225
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,32,2,power_law_1.01,1.8374336242675782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,32,2,power_law_1.01,0.1663884735107422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,32,2,power_law_1.01,0.17808639526367187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,32,2,power_law_1.01,0.2054195213317871
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,32,2,power_law_1.01,0.22314111709594728
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,32,2,power_law_1.01,0.2971225547790527
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,32,2,power_law_1.01,0.10561152458190919
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,32,2,power_law_1.01,0.3420915222167969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,32,2,power_law_1.01,0.43598590850830077
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,32,2,power_law_1.2,0.030154240131378175
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,32,2,power_law_1.01,0.5275532913208008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,32,2,power_law_1.2,0.04875775814056397
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,32,2,power_law_1.01,0.7506009674072265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,32,2,power_law_1.01,0.16456703186035154
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,32,2,power_law_1.2,0.0669478416442871
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,32,2,power_law_1.2,0.08087039947509765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,32,2,power_law_1.01,0.9257804870605468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,32,2,power_law_1.2,0.0848806381225586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,32,2,power_law_1.2,0.10854656219482421
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,32,2,power_law_1.2,0.11143808364868164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,32,2,power_law_1.2,0.12209664344787599
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,32,2,power_law_1.2,0.11161343574523926
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,32,2,power_law_1.01,1.3115724182128905
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,32,2,power_law_1.2,0.1538047981262207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,32,2,power_law_1.2,0.1603366470336914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,32,2,power_law_1.2,0.16312192916870116
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,32,2,power_law_1.2,0.16945663452148438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,32,2,power_law_1.2,0.18206592559814455
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,32,2,power_law_1.2,0.0562713623046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,32,2,power_law_1.2,0.20604288101196291
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,32,2,power_law_1.2,0.22463743209838868
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,32,2,power_law_1.2,0.1022719955444336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,32,2,power_law_1.2,0.3121062469482422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,32,2,power_law_1.2,0.39679489135742185
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,32,2,power_law_1.2,0.15202688217163085
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,32,2,power_law_1.2,0.5274470520019532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,32,2,power_law_1.2,0.03197695970535278
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,32,2,power_law_1.2,0.03641088008880615
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,32,2,power_law_1.01,1.7490098571777346
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,32,2,power_law_1.2,0.04487423896789551
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,32,2,power_law_1.2,0.055854082107543945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,32,2,power_law_1.2,0.26998912811279296
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,32,2,power_law_1.2,0.06858880043029784
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,32,2,power_law_1.2,0.0807795238494873
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,32,2,power_law_1.2,0.09905407905578614
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,32,2,power_law_1.2,0.10653568267822267
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,32,2,power_law_1.2,0.10547327995300293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,32,2,power_law_1.2,0.10847999572753905
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,32,2,power_law_1.2,0.1128166389465332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,32,2,power_law_1.2,0.7480025482177735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,32,2,power_law_1.2,0.1509990406036377
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,32,2,power_law_1.2,0.15680255889892578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,32,2,power_law_1.2,0.16282880783081055
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,32,2,power_law_1.2,0.16598783493041994
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,32,2,power_law_1.2,1.8566259765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,32,2,power_law_1.2,0.17069311141967775
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,32,2,power_law_1.2,0.18245887756347656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,32,2,power_law_1.2,1.0009535980224609
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,32,2,power_law_1.2,0.2549926376342774
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,32,2,power_law_1.2,0.3092787170410156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,32,2,power_law_1.2,0.351965446472168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,32,2,power_law_1.2,1.421939239501953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,32,2,power_law_1.2,0.4434009552001953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,32,2,power_law_1.2,0.5471270370483399
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,1,64,balanced,0.01789695978164673
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,1,64,balanced,0.017980159521102906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,1,64,balanced,0.017886719703674316
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,32,2,power_law_1.2,0.7506086730957031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,1,64,balanced,0.018018560409545897
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,1,64,balanced,0.024421119689941408
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,1,64,balanced,0.024931840896606446
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,1,64,balanced,0.024999680519104003
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,1,64,balanced,0.026794240474700925
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,32,2,power_law_1.2,0.23438976287841795
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,1,64,balanced,0.0266483211517334
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,1,64,balanced,0.027261440753936765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,1,64,balanced,0.029753599166870114
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,1,64,balanced,0.030314240455627438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,1,64,balanced,0.030744318962097165
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,1,64,balanced,0.03057152032852173
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,1,64,balanced,0.030703361034393313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,1,64,balanced,0.03262847900390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,1,64,balanced,0.0348198390007019
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,32,2,power_law_1.2,1.3902041625976562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,1,64,balanced,0.04144896030426025
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,1,64,balanced,0.04662399768829346
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,1,64,balanced,0.01783679962158203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,1,64,balanced,0.05377151966094971
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,1,64,balanced,0.06164608001708984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,1,64,balanced,0.07521535873413085
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,1,64,balanced,0.08638463973999023
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,1,64,balanced,0.12892800331115722
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,32,2,power_law_1.2,1.8165785217285155
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,1,64,balanced,0.021036798954010012
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,1,64,balanced,0.1588697624206543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,1,64,balanced,0.020716800689697265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,1,64,balanced,0.021172480583190916
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,1,64,balanced,0.030763518810272217
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,1,64,balanced,0.031223039627075195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,1,64,balanced,0.031123199462890626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,1,64,balanced,0.03153280019760132
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,1,64,balanced,0.03171200037002563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,1,64,balanced,0.03206912040710449
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,1,64,balanced,0.032289280891418456
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,1,64,balanced,0.03067392110824585
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,1,64,balanced,0.032660479545593264
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,1,64,balanced,0.0336409592628479
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,1,64,balanced,0.03435264110565185
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,1,64,balanced,0.03579904079437256
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,1,64,balanced,0.0382476806640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,1,64,balanced,0.03476736068725586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,1,64,balanced,0.03724800109863281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,1,64,balanced,0.04165247917175293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,1,64,balanced,0.04781824111938476
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,1,64,balanced,0.02062335968017578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,1,64,balanced,0.05265151977539062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,32,2,power_law_1.2,0.9638540649414062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,1,64,balanced,0.06286208152770996
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,1,64,balanced,0.07573631763458252
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,1,64,balanced,0.10072064399719238
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,1,64,balanced,0.12661120414733887
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,1,64,power_law_1.01,0.018346240520477293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,1,64,power_law_1.01,0.01807487964630127
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,1,64,power_law_1.01,0.01824512004852295
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,1,64,power_law_1.01,0.025248000621795653
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,1,64,power_law_1.01,0.024564480781555174
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,1,64,power_law_1.01,0.0266431999206543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,1,64,power_law_1.01,0.027423999309539794
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,1,64,power_law_1.01,0.031516160964965824
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,1,64,power_law_1.01,0.030827519893646238
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,1,64,power_law_1.01,0.03192575931549072
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,1,64,power_law_1.01,0.0326310396194458
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,1,64,power_law_1.01,0.034777600765228275
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,1,64,power_law_1.01,0.03751424074172974
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,1,64,power_law_1.01,0.04198016166687012
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,1,64,power_law_1.01,0.04744192123413086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,1,64,power_law_1.01,0.044477438926696776
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,1,64,power_law_1.01,0.05139328002929687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,1,64,power_law_1.01,0.06205056190490723
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,1,64,power_law_1.01,0.08262784004211425
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,1,64,power_law_1.01,0.10515071868896483
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,1,64,power_law_1.01,0.14908672332763673
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,1,64,balanced,0.020869119167327883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,1,64,power_law_1.01,0.18486143112182618
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,1,64,power_law_1.01,0.2711283111572266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,1,64,power_law_1.01,0.35834369659423826
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,1,64,power_law_1.01,0.021000959873199464
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,1,64,power_law_1.01,0.02098304033279419
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,1,64,power_law_1.01,0.03071104049682617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,1,64,power_law_1.01,0.030690560340881346
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,1,64,power_law_1.01,0.5671155166625976
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,1,64,power_law_1.01,0.031217920780181884
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,1,64,power_law_1.01,0.0322329592704773
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,1,64,power_law_1.01,0.7048537445068359
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,1,64,power_law_1.01,0.031973121166229246
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,1,64,power_law_1.01,0.03285504102706909
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,1,64,power_law_1.01,0.033180160522460936
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,1,64,power_law_1.01,0.034423038959503174
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,1,64,power_law_1.01,0.03598592042922973
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,1,64,power_law_1.01,0.037088000774383546
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,1,64,power_law_1.01,0.03878911972045899
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,1,64,power_law_1.01,0.037131519317626954
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,1,64,power_law_1.01,0.03864576101303101
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,1,64,power_law_1.01,0.0414796781539917
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,1,64,power_law_1.01,0.046489601135253904
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,1,64,power_law_1.01,0.05335807800292969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,1,64,power_law_1.01,0.06366975784301758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,1,64,power_law_1.01,0.07588863849639893
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,1,64,power_law_1.01,0.1096448040008545
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,1,64,power_law_1.01,0.021274878978729247
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,1,64,power_law_1.01,0.1255679988861084
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,1,64,power_law_1.01,0.1862553596496582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,1,64,power_law_1.01,0.26308479309082033
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,1,64,balanced,0.03350656032562256
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,1,64,power_law_1.2,0.018391040563583375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,1,64,power_law_1.2,0.02446079969406128
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,1,64,power_law_1.01,0.4162892913818359
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,1,64,power_law_1.01,0.03314176082611084
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,1,64,power_law_1.2,0.02474368095397949
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,1,64,power_law_1.2,0.025612800121307372
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,1,64,power_law_1.2,0.026708478927612307
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,1,64,power_law_1.01,0.5563238525390626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,1,64,power_law_1.2,0.02884864091873169
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,1,64,power_law_1.2,0.028847360610961915
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,1,64,power_law_1.2,0.03456640005111694
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,1,64,power_law_1.2,0.03684864044189453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,1,64,power_law_1.2,0.04457344055175781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,1,64,power_law_1.2,0.04491392135620117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,1,64,power_law_1.2,0.052390398979187014
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,1,64,power_law_1.2,0.060842242240905765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,1,64,power_law_1.2,0.018056960105895997
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,1,64,power_law_1.2,0.0764518404006958
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,1,64,power_law_1.2,0.11017344474792481
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,1,64,power_law_1.2,0.1334425640106201
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,1,64,power_law_1.2,0.0306329607963562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,1,64,power_law_1.2,0.030617599487304685
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,1,64,power_law_1.2,0.27658111572265626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,1,64,power_law_1.2,0.033858559131622314
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,1,64,power_law_1.2,0.020665600299835205
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,1,64,power_law_1.2,0.021329920291900634
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,1,64,power_law_1.2,0.030869760513305665
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,1,64,power_law_1.2,0.032647678852081294
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,1,64,power_law_1.2,0.031203839778900143
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,1,64,power_law_1.2,0.031560959815979006
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,1,64,power_law_1.2,0.03213696002960205
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,1,64,power_law_1.2,0.3879884719848633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,1,64,power_law_1.2,0.03237375974655151
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,1,64,power_law_1.2,0.03318144083023071
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,1,64,power_law_1.2,0.03357952117919922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,1,64,power_law_1.2,0.03376768112182617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,1,64,power_law_1.2,0.03695487976074219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,1,64,power_law_1.2,0.03718528032302856
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,1,64,power_law_1.2,0.03889152050018311
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,1,64,power_law_1.2,0.037414400577545165
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,1,64,power_law_1.2,0.19869951248168943
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,1,64,power_law_1.2,0.042175998687744146
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,1,64,power_law_1.2,0.051072001457214355
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,1,64,power_law_1.2,0.06030079841613769
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,1,64,power_law_1.2,0.07164927959442138
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,1,64,power_law_1.2,0.09754112243652344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,1,64,power_law_1.2,0.16481920242309572
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,1,64,power_law_1.2,0.035848960876464844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,1,64,power_law_1.2,0.30825855255126955
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,1,64,power_law_1.2,0.19361919403076172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,1,64,power_law_1.2,0.038279678821563724
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,2,32,balanced,0.01649664044380188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,2,32,balanced,0.016441600322723387
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,2,32,balanced,0.01660159945487976
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,2,32,balanced,0.016892160177230837
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,2,32,balanced,0.017824000120162962
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,2,32,balanced,0.023576319217681885
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,2,32,balanced,0.024194560050964355
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,2,32,balanced,0.024567039012908937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,2,32,balanced,0.025251839160919186
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,2,32,balanced,0.025269761085510257
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,2,32,balanced,0.0262719988822937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,2,32,balanced,0.02730623960494995
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,2,32,balanced,0.028209919929504397
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,2,32,balanced,0.030970880985260008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,2,32,balanced,0.032376320362091066
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,2,32,balanced,0.029671680927276612
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,2,32,balanced,0.03541376113891602
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,2,32,balanced,0.0419814395904541
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,2,32,balanced,0.04855040073394775
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,2,32,balanced,0.05493760108947754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,2,32,balanced,0.06739967823028564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,2,32,balanced,0.03221375942230224
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,2,32,balanced,0.09950336456298828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,2,32,balanced,0.12029312133789063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,1,64,power_law_1.2,0.5228377532958984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,2,32,balanced,0.17566591262817383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,1,64,power_law_1.2,0.6150912094116211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,2,32,balanced,0.22279680252075193
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,2,32,balanced,0.01969023942947388
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,2,32,balanced,0.01965824007987976
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,2,32,balanced,0.019701759815216064
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,2,32,balanced,0.019838720560073853
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,2,32,balanced,0.020200960636138916
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,2,32,balanced,0.07640063762664795
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,2,32,balanced,0.029407999515533446
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,2,32,balanced,0.02974976062774658
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,2,32,balanced,0.03010175943374634
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,2,32,balanced,0.030716159343719483
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,2,32,balanced,0.03169152021408081
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,2,32,balanced,0.03177855968475342
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,2,32,balanced,0.032671999931335446
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,2,32,balanced,0.033575680255889896
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,2,32,balanced,0.034757120609283446
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,2,32,balanced,0.035861759185791015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,2,32,balanced,0.03216255903244018
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,2,32,balanced,0.03449856042861939
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,2,32,balanced,0.037529599666595456
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,2,32,balanced,0.04193920135498047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,2,32,balanced,0.04822527885437012
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,2,32,balanced,0.05852159976959228
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,2,32,balanced,0.0679040002822876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,2,32,balanced,0.08995455741882324
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,2,32,balanced,0.11225215911865234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,2,32,balanced,0.15345024108886718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,2,32,balanced,0.030849280357360842
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,2,32,balanced,0.1919206428527832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,2,32,power_law_1.01,0.017455359697341917
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,1,64,power_law_1.2,0.3916223907470703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,2,32,power_law_1.01,0.0174233603477478
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,2,32,power_law_1.01,0.017822719812393188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,2,32,power_law_1.01,0.023449599742889404
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,2,32,power_law_1.01,0.02416127920150757
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,2,32,power_law_1.01,0.0254694390296936
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,2,32,power_law_1.01,0.02673151969909668
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,2,32,power_law_1.01,0.026338560581207277
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,2,32,power_law_1.01,0.02861311912536621
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,2,32,power_law_1.01,0.031157760620117186
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,2,32,power_law_1.01,0.03237119913101196
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,2,32,power_law_1.01,0.033831679821014406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,2,32,power_law_1.01,0.03335936069488525
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,2,32,power_law_1.01,0.038060801029205324
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,2,32,power_law_1.01,0.04156032085418701
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,2,32,power_law_1.01,0.04679296016693115
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,2,32,power_law_1.01,0.05567615985870361
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,2,32,power_law_1.01,0.06149888038635254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,2,32,power_law_1.01,0.07456768035888672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,2,32,power_law_1.01,0.08988287925720215
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,2,32,power_law_1.01,0.024527359008789062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,2,32,power_law_1.01,0.1235148811340332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,2,32,power_law_1.01,0.17166208267211913
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,2,32,power_law_1.01,0.2208460807800293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,2,32,power_law_1.01,0.35635326385498045
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,1,64,power_law_1.2,0.8155059051513671
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,2,32,power_law_1.01,0.02002432107925415
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,2,32,power_law_1.01,0.4881011199951172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,2,32,power_law_1.01,0.02027008056640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,2,32,power_law_1.01,0.029255681037902832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,2,32,power_law_1.01,0.6113292694091796
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,2,32,power_law_1.01,0.029747200012207032
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,2,32,power_law_1.01,0.030099198818206784
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,2,32,power_law_1.01,0.03146752119064331
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,2,32,power_law_1.01,0.03192575931549072
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,2,32,power_law_1.01,0.03348223924636841
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,2,32,power_law_1.01,0.0346560001373291
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,2,32,power_law_1.01,0.03664767980575562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,2,32,power_law_1.01,0.03546367883682251
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,2,32,power_law_1.01,0.03499520063400269
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,2,32,power_law_1.01,0.03726847887039185
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,2,32,power_law_1.01,0.0408025598526001
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,2,32,power_law_1.01,0.04658048152923584
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,1,64,power_law_1.2,1.0709414672851563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,2,32,power_law_1.01,0.05101823806762695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,2,32,power_law_1.01,0.05983359813690185
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,1,64,power_law_1.2,0.8226879882812501
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,2,32,power_law_1.01,0.02012160062789917
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,2,32,power_law_1.01,0.10291839599609376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,2,32,power_law_1.01,0.14070528030395507
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,2,32,power_law_1.01,0.030812160968780516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,2,32,power_law_1.01,0.032057600021362306
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,2,32,power_law_1.01,0.2071615982055664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,2,32,power_law_1.2,0.017360639572143555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,2,32,power_law_1.01,0.2719193649291992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,2,32,power_law_1.2,0.01757696032524109
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,2,32,power_law_1.01,0.396866569519043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,2,32,power_law_1.2,0.023365120887756347
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,2,32,power_law_1.2,0.025052158832550048
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,2,32,power_law_1.2,0.025598719120025638
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,2,32,power_law_1.01,0.5364710235595703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,2,32,power_law_1.2,0.025652480125427247
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,2,32,power_law_1.2,0.027340800762176515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,2,32,power_law_1.2,0.028843519687652586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,2,32,power_law_1.2,0.03219072103500366
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,2,32,power_law_1.2,0.03273983955383301
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,2,32,power_law_1.2,0.03991552114486695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,2,32,power_law_1.2,0.043854079246520995
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,2,32,power_law_1.2,0.04398719787597656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,2,32,power_law_1.01,0.0816140842437744
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,2,32,power_law_1.2,0.04846464157104492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,2,32,power_law_1.2,0.05926655769348145
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,2,32,power_law_1.2,0.06952576160430908
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,2,32,power_law_1.2,0.08175999641418456
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,2,32,power_law_1.2,0.1061747169494629
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,2,32,power_law_1.2,0.1674163246154785
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,2,32,power_law_1.2,0.01779584050178528
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,2,32,power_law_1.2,0.02398848056793213
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,2,32,power_law_1.2,0.019943679571151736
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,2,32,power_law_1.2,0.020179200172424316
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,2,32,power_law_1.2,0.03355263948440552
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,2,32,power_law_1.2,0.020367360115051268
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,2,32,power_law_1.2,0.02926464080810547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,2,32,power_law_1.2,0.22716415405273435
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,2,32,power_law_1.2,0.029726719856262206
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,2,32,power_law_1.2,0.03029632091522217
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,2,32,power_law_1.2,0.030996479988098145
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,2,32,power_law_1.2,0.03173120021820068
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,2,32,power_law_1.2,0.032065279483795166
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,2,32,power_law_1.2,0.032440319061279296
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,2,32,power_law_1.2,0.03478015899658203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,2,32,power_law_1.2,0.03684096097946167
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,2,32,power_law_1.2,0.03579391956329346
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,2,32,power_law_1.2,0.03447168111801148
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,2,32,power_law_1.2,0.03663872003555298
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,2,32,power_law_1.2,0.03839744091033935
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,2,32,power_law_1.2,0.04298367977142334
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,2,32,power_law_1.2,0.05079552173614502
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,2,32,power_law_1.2,0.05785727977752686
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,2,32,power_law_1.2,0.33711360931396483
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,2,32,power_law_1.2,0.07433599948883056
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,2,32,power_law_1.2,0.09058815956115722
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,2,32,power_law_1.2,0.11658623695373535
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,2,32,power_law_1.2,0.18907392501831055
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,2,32,power_law_1.2,0.25682687759399414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,4,16,balanced,0.016302080154418946
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,2,32,power_law_1.2,0.5084672164916992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,4,16,balanced,0.01651455998420715
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,4,16,balanced,0.017311999797821047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,4,16,balanced,0.01871616005897522
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,4,16,balanced,0.024216320514678955
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,4,16,balanced,0.025566720962524415
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,4,16,balanced,0.02645632028579712
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,4,16,balanced,0.027648000717163085
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,4,16,balanced,0.027714560031890868
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,4,16,balanced,0.029470720291137696
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,4,16,balanced,0.031169281005859372
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,4,16,balanced,0.026621439456939698
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,4,16,balanced,0.01678976058959961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,4,16,balanced,0.03145344018936157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,4,16,balanced,0.03705087900161743
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,4,16,balanced,0.04072959899902344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,4,16,balanced,0.044405760765075686
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,4,16,balanced,0.026795520782470706
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,4,16,balanced,0.052203521728515626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,4,16,balanced,0.06375296115875244
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,2,32,power_law_1.2,0.35593471527099607
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,4,16,balanced,0.07189504146575929
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,4,16,balanced,0.09195263862609863
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,2,32,power_law_1.2,0.44036609649658204
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,4,16,balanced,0.1095961570739746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,4,16,balanced,0.1508838367462158
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,4,16,balanced,0.18842111587524416
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,4,16,balanced,0.019920639991760254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,4,16,balanced,0.019674880504608156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,4,16,balanced,0.019992320537567138
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,4,16,balanced,0.35185665130615235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,4,16,balanced,0.021870079040527343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,4,16,balanced,0.030388479232788083
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,4,16,balanced,0.030833919048309327
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,4,16,balanced,0.031516160964965824
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,4,16,balanced,0.03172352075576782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,4,16,balanced,0.03221888065338135
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,4,16,balanced,0.0334771203994751
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,4,16,balanced,0.0347161602973938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,4,16,balanced,0.03574656009674072
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,4,16,balanced,0.03173120021820068
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,4,16,balanced,0.034197759628295896
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,4,16,balanced,0.03506688117980957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,4,16,balanced,0.036686079502105715
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,4,16,balanced,0.04000256061553955
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,4,16,balanced,0.04662144184112549
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,4,16,balanced,0.05461887836456299
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,4,16,balanced,0.27003007888793945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,4,16,balanced,0.06444543838500977
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,4,16,balanced,0.08715519905090333
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,4,16,balanced,0.02040447950363159
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,4,16,balanced,0.10581760406494141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,4,16,balanced,0.14325119972229003
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,4,16,balanced,0.17979520797729492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,4,16,power_law_1.01,0.017473280429840088
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,4,16,balanced,0.2578790473937988
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,4,16,power_law_1.01,0.017648639678955077
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,4,16,power_law_1.01,0.018165760040283203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,4,16,balanced,0.3285516738891602
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,4,16,power_law_1.01,0.01842687964439392
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,4,16,power_law_1.01,0.023878400325775144
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,4,16,power_law_1.01,0.024748799800872804
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,4,16,power_law_1.01,0.0263321590423584
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,4,16,power_law_1.01,0.02696448087692261
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,4,16,power_law_1.01,0.027439360618591306
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,4,16,power_law_1.01,0.02806528091430664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,4,16,power_law_1.01,0.03315711975097656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,4,16,power_law_1.01,0.032883200645446774
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,4,16,power_law_1.01,0.03595135927200317
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,4,16,power_law_1.01,0.033637120723724365
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,4,16,power_law_1.01,0.03940479993820191
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,4,16,power_law_1.01,0.041840639114379886
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,4,16,power_law_1.01,0.04632832050323486
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,4,16,power_law_1.01,0.05759615898132324
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,4,16,power_law_1.01,0.0645363187789917
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,4,16,power_law_1.01,0.07558527946472168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,4,16,power_law_1.01,0.08791423797607421
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,4,16,power_law_1.01,0.12031488418579102
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,4,16,power_law_1.01,0.1460223960876465
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,4,16,power_law_1.01,0.23519872665405273
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,4,16,power_law_1.01,0.02027776002883911
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,4,16,power_law_1.01,0.33234302520751957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,4,16,power_law_1.01,0.020753920078277588
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,4,16,power_law_1.01,0.4452313613891602
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,4,16,power_law_1.01,0.02144511938095093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,4,16,power_law_1.01,0.021850879192352297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,4,16,power_law_1.01,0.6044275283813476
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,4,16,power_law_1.01,0.02993664026260376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,4,16,power_law_1.01,0.030538239479064942
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,4,16,power_law_1.01,0.0318016004562378
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,4,16,power_law_1.01,0.032951040267944334
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,4,16,power_law_1.01,0.032975358963012694
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,4,16,power_law_1.01,0.03434880018234253
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,4,16,power_law_1.01,0.03668864011764526
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,4,16,power_law_1.01,0.03606784105300903
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,4,16,power_law_1.01,0.03470976114273071
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,4,16,power_law_1.01,0.0370739197731018
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,4,16,power_law_1.01,0.03863168001174927
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,4,16,power_law_1.01,0.04271999835968018
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,4,16,power_law_1.01,0.048220157623291016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,4,16,power_law_1.01,0.05661439895629883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,4,16,power_law_1.01,0.06821760177612304
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,4,16,power_law_1.01,0.08687616348266601
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,4,16,power_law_1.01,0.03137279987335205
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,4,16,power_law_1.01,0.11281279563903808
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,2,32,power_law_1.2,0.6670259094238281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,4,16,power_law_1.01,0.1440460777282715
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,4,16,power_law_1.01,0.2154636764526367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,4,16,power_law_1.01,0.2618675231933594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,4,16,power_law_1.01,0.394598388671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,4,16,power_law_1.2,0.017845760583877563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,4,16,power_law_1.2,0.018136320114135743
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,4,16,power_law_1.01,0.5489523315429687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,4,16,power_law_1.2,0.01841920018196106
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,4,16,power_law_1.2,0.023400959968566896
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,4,16,power_law_1.2,0.02504192113876343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,4,16,power_law_1.2,0.027022080421447752
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,4,16,power_law_1.2,0.026077439785003664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,4,16,power_law_1.2,0.02815104007720947
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,4,16,power_law_1.2,0.031517438888549805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,4,16,power_law_1.2,0.03500159978866577
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,4,16,power_law_1.2,0.03659136056900024
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,4,16,power_law_1.2,0.03880575895309448
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,4,16,power_law_1.2,0.0442739200592041
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,4,16,power_law_1.2,0.01750272035598755
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,4,16,power_law_1.2,0.05059455871582032
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,4,16,power_law_1.2,0.0584012794494629
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,4,16,power_law_1.2,0.06757887840270996
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,4,16,power_law_1.2,0.07999104022979736
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,4,16,power_law_1.2,0.10404095649719239
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,4,16,power_law_1.2,0.14941823959350586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,4,16,power_law_1.2,0.027408640384674072
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,4,16,power_law_1.2,0.18077568054199217
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,4,16,power_law_1.2,0.020225279331207276
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,4,16,power_law_1.2,0.021245439052581788
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,4,16,power_law_1.2,0.02144000053405762
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,4,16,power_law_1.2,0.021475839614868163
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,4,16,power_law_1.2,0.02981760025024414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,4,16,power_law_1.2,0.041301760673522946
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,4,16,power_law_1.2,0.03058687925338745
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,4,16,power_law_1.2,0.03127039909362793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,4,16,power_law_1.2,0.031784958839416504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,4,16,power_law_1.2,0.03278336048126221
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,4,16,power_law_1.2,0.033571839332580566
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,4,16,power_law_1.2,0.03551232099533081
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,4,16,power_law_1.2,0.03588479995727539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,4,16,power_law_1.2,0.03394815921783447
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,4,16,power_law_1.2,0.03571840047836304
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,4,16,power_law_1.2,0.03784192085266113
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,4,16,power_law_1.2,0.03979135990142822
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,4,16,power_law_1.2,0.04446976184844971
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,4,16,power_law_1.2,0.051466240882873535
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,4,16,power_law_1.2,0.0612774419784546
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,4,16,power_law_1.2,0.07274879932403563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,4,16,power_law_1.2,0.09527039527893066
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,4,16,power_law_1.2,0.16939647674560548
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,4,16,power_law_1.2,0.13979392051696776
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,2,32,power_law_1.2,0.827545623779297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,4,16,power_law_1.2,0.25438720703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,4,16,power_law_1.2,0.34641662597656253
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,1,64,balanced,0.025480959415435787
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,1,64,balanced,0.025633280277252195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,1,64,balanced,0.025288960933685302
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,4,16,power_law_1.2,0.2895193672180176
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,1,64,balanced,0.02550271987915039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,1,64,balanced,0.04921088218688965
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,1,64,balanced,0.08249600410461426
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,1,64,balanced,0.08296575546264648
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,1,64,balanced,0.08341119766235353
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,1,64,balanced,0.08385663986206054
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,1,64,balanced,0.08362879753112792
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,4,16,power_law_1.2,0.3157414436340332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,1,64,balanced,0.08559743881225586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,1,64,balanced,0.08733440399169921
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,1,64,balanced,0.08813183784484864
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,1,64,balanced,0.0909273624420166
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,1,64,balanced,0.09196160316467286
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,1,64,balanced,0.09611519813537597
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,1,64,balanced,0.10045696258544923
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,4,16,power_law_1.2,0.547320327758789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,1,64,balanced,0.09606016159057618
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,1,64,balanced,0.10347392082214354
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,1,64,balanced,0.10731391906738282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,1,64,balanced,0.12018943786621095
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,1,64,balanced,0.1312384033203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,1,64,balanced,0.1876915168762207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,1,64,balanced,0.23169151306152344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,1,64,balanced,0.08870143890380859
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,1,64,balanced,0.027095038890838623
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,1,64,balanced,0.4137855911254883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,1,64,balanced,0.027098879814147946
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,1,64,balanced,0.02712064027786255
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,1,64,balanced,0.027795200347900388
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,1,64,balanced,0.04411776065826416
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,2,32,power_law_1.2,0.6611315155029297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,1,64,balanced,0.04498047828674316
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,1,64,balanced,0.04628608226776123
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,1,64,balanced,0.3250073623657227
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,1,64,balanced,0.04486144065856933
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,1,64,balanced,0.046846718788146974
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,1,64,balanced,0.04601856231689453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,1,64,balanced,0.04627327919006348
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,1,64,balanced,0.0519052791595459
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,1,64,balanced,0.04754432201385498
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,1,64,balanced,0.05105792045593262
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,1,64,balanced,0.052614398002624505
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,1,64,balanced,0.06033408164978027
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,1,64,balanced,0.05536255836486816
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,1,64,balanced,0.06156928062438964
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,1,64,balanced,0.06703999996185303
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,1,64,balanced,0.07502592086791993
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,1,64,balanced,0.08450816154479981
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,1,64,balanced,0.10458751678466797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,1,64,balanced,0.13166848182678223
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,1,64,balanced,0.026684160232543948
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,1,64,balanced,0.04829184055328369
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,1,64,balanced,0.18813056945800782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,1,64,power_law_1.01,0.04924160003662109
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,1,64,power_law_1.01,0.04918911933898926
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,1,64,power_law_1.01,0.08195967674255371
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,1,64,power_law_1.01,0.08343680381774903
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,1,64,power_law_1.01,0.08424192428588867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,1,64,power_law_1.01,0.08668160438537598
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,1,64,power_law_1.01,0.08715007781982423
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,1,64,power_law_1.01,0.08890239715576172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,1,64,power_law_1.01,0.09024127960205078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,4,16,power_law_1.2,0.4898112106323242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,1,64,power_law_1.01,0.09361023902893066
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,1,64,power_law_1.01,0.09754752159118653
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,1,64,power_law_1.01,0.10054400444030762
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,1,64,power_law_1.01,0.10132479667663574
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,1,64,balanced,0.23052160263061525
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,1,64,power_law_1.01,0.11378944396972657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,1,64,power_law_1.01,0.1322124767303467
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,1,64,power_law_1.01,0.04935296058654785
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,1,64,power_law_1.01,0.1556492805480957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,1,64,power_law_1.01,0.2192486381530762
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,1,64,power_law_1.01,0.2863615989685059
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,1,64,power_law_1.01,0.4191782379150391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,1,64,power_law_1.01,0.5321382522583008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,1,64,power_law_1.01,0.10320128440856932
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,1,64,power_law_1.01,0.10500608444213869
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,1,64,power_law_1.01,0.027701759338378908
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,1,64,power_law_1.01,0.8541670227050782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,1,64,power_law_1.01,0.029151999950408937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,1,64,power_law_1.01,0.02739840030670166
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,1,64,power_law_1.01,0.04410240173339844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,1,64,power_law_1.01,0.044568319320678715
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,1,64,power_law_1.01,1.1486605072021485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,1,64,power_law_1.01,0.04566527843475342
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,1,64,power_law_1.01,0.045646080970764155
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,1,64,power_law_1.01,0.047491841316223145
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,1,64,power_law_1.01,0.04834559917449951
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,1,64,power_law_1.01,0.049538559913635254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,1,64,power_law_1.01,0.05212927818298339
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,1,64,power_law_1.01,0.05276927947998047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,1,64,power_law_1.01,0.05803520202636718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,1,64,power_law_1.01,0.059388160705566406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,1,64,power_law_1.01,0.06058879852294922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,1,64,power_law_1.01,0.06015999794006348
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,1,64,power_law_1.01,0.06345600128173828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,1,64,power_law_1.01,0.07479423999786376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,1,64,power_law_1.01,0.0861888027191162
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,1,64,power_law_1.01,0.09944576263427735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,1,64,power_law_1.01,0.12485504150390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,1,64,power_law_1.01,0.16723199844360354
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,1,64,power_law_1.01,0.2484364891052246
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,1,64,power_law_1.01,2.425605163574219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,1,64,power_law_1.01,0.41522689819335934
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,1,64,power_law_1.2,0.04902527809143066
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,1,64,power_law_1.2,0.04921728134155273
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,1,64,power_law_1.01,1.869178924560547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,1,64,power_law_1.01,0.7022796630859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,1,64,power_law_1.2,0.08210944175720215
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,1,64,power_law_1.2,0.08218624114990233
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,1,64,power_law_1.2,0.08364928245544434
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,1,64,power_law_1.2,0.08422143936157227
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,1,64,power_law_1.2,0.08834431648254396
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,1,64,power_law_1.01,0.9715699005126954
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,1,64,power_law_1.2,0.09050623893737793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,1,64,power_law_1.2,0.09128576278686523
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,1,64,power_law_1.2,0.09775615692138671
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,1,64,power_law_1.2,0.09949952125549316
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,1,64,power_law_1.2,0.10363264083862304
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,1,64,power_law_1.2,0.10242431640625001
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,1,64,power_law_1.2,0.10059391975402832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,1,64,power_law_1.2,0.10963711738586426
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,1,64,power_law_1.01,0.5644863891601563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,1,64,power_law_1.2,0.12217856407165528
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,1,64,power_law_1.2,0.16056320190429688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,1,64,power_law_1.2,0.2002649688720703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,1,64,power_law_1.2,0.2900313568115235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,1,64,power_law_1.2,0.4040537643432617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,1,64,power_law_1.2,0.08794495582580567
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,1,64,power_law_1.2,0.6411251068115235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,1,64,power_law_1.2,0.8797196960449218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,1,64,power_law_1.2,0.027482879161834717
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,1,64,power_law_1.2,0.02986112117767334
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,1,64,power_law_1.2,0.029573121070861817
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,1,64,power_law_1.2,0.04491136074066162
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,4,16,power_law_1.2,0.7095053100585937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,1,64,power_law_1.2,0.04614783763885498
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,1,64,power_law_1.2,0.04564352035522461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,1,64,power_law_1.2,0.049203200340270994
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,1,64,power_law_1.2,0.05047935962677001
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,1,64,power_law_1.2,0.05074175834655762
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,1,64,power_law_1.2,0.048712959289550786
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,1,64,power_law_1.2,0.05416192054748535
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,1,64,power_law_1.2,0.056412158012390134
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,1,64,power_law_1.2,0.05938047885894775
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,1,64,power_law_1.2,0.062417922019958494
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,1,64,power_law_1.2,0.060189437866210935
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,1,64,power_law_1.2,0.06240511894226074
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,1,64,power_law_1.2,0.06756608009338379
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,1,64,power_law_1.2,0.08341119766235353
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,1,64,power_law_1.2,0.0985804843902588
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,1,64,power_law_1.2,0.1372390365600586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,1,64,power_law_1.2,0.2166758346557617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,1,64,power_law_1.2,0.2965043258666992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,1,64,power_law_1.2,0.37445888519287107
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,1,64,power_law_1.2,1.6276441955566405
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,4,16,power_law_1.2,0.6387020874023437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,2,32,balanced,0.02325119972229004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,1,64,power_law_1.2,1.2565299224853514
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,2,32,balanced,0.023182079792022706
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,2,32,balanced,0.02312448024749756
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,2,32,balanced,0.03778048038482666
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,2,32,balanced,0.04591104030609131
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,2,32,balanced,0.07485951900482177
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,2,32,balanced,0.0767244815826416
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,2,32,balanced,0.07705984115600586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,2,32,balanced,0.07817728042602538
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,2,32,balanced,0.07912703990936279
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,2,32,balanced,0.07994112014770507
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,2,32,balanced,0.08213248252868652
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,2,32,balanced,0.08444031715393066
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,2,32,balanced,0.08601216316223145
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,2,32,balanced,0.09011072158813477
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,2,32,balanced,0.09142399787902832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,2,32,balanced,0.0869644832611084
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,2,32,balanced,0.09239295959472657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,2,32,balanced,0.09550848007202148
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,2,32,balanced,0.1051865577697754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,2,32,balanced,0.11319168090820313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,2,32,balanced,0.1350220775604248
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,2,32,balanced,0.1523532772064209
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,2,32,balanced,0.22189823150634766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,2,32,balanced,0.2797196769714355
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,2,32,balanced,0.39828609466552733
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,2,32,balanced,0.02482304096221924
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,2,32,balanced,0.5063859176635742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,2,32,balanced,0.024738559722900393
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,2,32,balanced,0.024748799800872804
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,2,32,balanced,0.025340158939361573
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,2,32,balanced,0.027705600261688234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,2,32,balanced,0.04068607807159424
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,2,32,balanced,0.04181375980377197
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,2,32,balanced,0.04261248111724854
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,2,32,balanced,0.04314239978790283
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,2,32,balanced,0.043359999656677244
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,2,32,balanced,0.044558081626892086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,2,32,balanced,0.04494463920593262
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,2,32,balanced,0.047109122276306155
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,2,32,balanced,0.049292798042297366
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,2,32,balanced,0.054248957633972174
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,2,32,balanced,0.05703680038452148
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,1,64,power_law_1.2,0.6684774780273438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,2,32,balanced,0.05068543910980224
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,2,32,balanced,0.05557631969451904
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,2,32,balanced,0.06018176078796387
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,2,32,balanced,0.06634880065917968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,2,32,balanced,0.07433472156524659
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,2,32,balanced,0.09702143669128419
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,2,32,balanced,0.11212160110473632
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,2,32,balanced,0.14623488426208495
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,2,32,balanced,0.18235904693603516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,2,32,balanced,0.26989824295043946
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,2,32,power_law_1.01,0.04565887928009033
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,2,32,balanced,0.3354969787597656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,2,32,power_law_1.01,0.04564095973968506
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,2,32,power_law_1.01,0.046242561340332035
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,2,32,power_law_1.01,0.07276288032531739
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,2,32,power_law_1.01,0.07499775886535645
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,2,32,power_law_1.01,0.07837823867797852
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,2,32,power_law_1.01,0.07872767925262451
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,2,32,power_law_1.01,0.07963903903961181
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,1,64,power_law_1.2,2.6009204101562498
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,2,32,power_law_1.01,0.08125311851501464
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,2,32,power_law_1.01,0.08268287658691406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,2,32,power_law_1.01,0.0865459156036377
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,2,32,power_law_1.01,0.08951807975769042
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,2,32,power_law_1.01,0.09050623893737793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,2,32,power_law_1.01,0.09265151977539063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,2,32,power_law_1.01,0.09176704406738281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,2,32,power_law_1.01,0.09426815986633301
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,2,32,power_law_1.01,0.10341504096984863
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,2,32,power_law_1.01,0.11410688400268554
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,2,32,power_law_1.01,0.13052800178527832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,2,32,power_law_1.01,0.16900352478027342
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,2,32,power_law_1.01,0.23620351791381836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,2,32,power_law_1.01,0.32194942474365235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,1,64,power_law_1.2,0.851352310180664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,2,32,power_law_1.01,0.42131839752197264
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,2,32,power_law_1.01,0.6058380889892578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,2,32,power_law_1.01,0.02745471954345703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,2,32,power_law_1.01,0.8157299041748047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,2,32,power_law_1.01,0.0272870397567749
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,2,32,power_law_1.01,0.027703039646148682
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,2,32,power_law_1.01,1.1465984344482423
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,2,32,power_law_1.01,0.04067071914672852
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,2,32,power_law_1.01,0.04132863998413086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,2,32,power_law_1.01,0.04214655876159668
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,2,32,power_law_1.01,0.04424960136413574
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,2,32,power_law_1.01,0.04406144142150879
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,2,32,power_law_1.01,0.04910336017608642
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,2,32,power_law_1.01,0.04874112129211426
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,2,32,power_law_1.01,0.052110080718994145
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,2,32,power_law_1.01,0.053344001770019536
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,2,32,power_law_1.01,0.056894721984863283
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,2,32,power_law_1.01,0.05363711833953857
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,2,32,power_law_1.01,0.05693823814392089
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,2,32,power_law_1.01,0.05900288105010987
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,2,32,power_law_1.01,0.06169472217559815
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,2,32,power_law_1.01,0.07170815944671631
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,2,32,power_law_1.01,0.08321023941040039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,2,32,power_law_1.01,0.10791935920715331
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,2,32,power_law_1.01,0.1308415985107422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,2,32,power_law_1.01,0.19116416931152344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,2,32,power_law_1.01,0.2556211280822754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,2,32,power_law_1.01,1.6268902587890626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,2,32,power_law_1.01,0.40677761077880864
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,2,32,power_law_1.2,0.04705535888671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,2,32,power_law_1.2,0.0458457612991333
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,2,32,power_law_1.01,0.6839167785644531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,2,32,power_law_1.2,0.04442495822906494
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,2,32,power_law_1.2,0.07307392120361328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,2,32,power_law_1.2,0.07714431762695313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,2,32,power_law_1.01,0.9451136016845704
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,2,32,power_law_1.2,0.08142208099365235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,2,32,power_law_1.2,0.08113408088684082
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,2,32,power_law_1.01,0.3275225448608398
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,2,32,power_law_1.2,0.08402815818786621
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,2,32,power_law_1.2,0.08512255668640137
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,2,32,power_law_1.2,0.08966527938842773
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,2,32,power_law_1.2,0.09360768318176269
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,2,32,power_law_1.2,0.09228416442871093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,2,32,power_law_1.2,0.09349887847900391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,2,32,power_law_1.2,0.10204031944274902
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,2,32,power_law_1.2,0.10627072334289552
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,2,32,power_law_1.2,0.13279359817504882
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,2,32,power_law_1.2,0.0756608009338379
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,2,32,power_law_1.2,0.2129523277282715
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,2,32,power_law_1.2,0.3271027374267578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,2,32,power_law_1.2,0.09018367767333985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,1,64,power_law_1.2,1.128870391845703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,2,32,power_law_1.2,0.4540748977661133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,2,32,power_law_1.2,0.5518476867675781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,2,32,power_law_1.2,0.15849087715148927
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,2,32,power_law_1.2,0.02728192090988159
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,2,32,power_law_1.2,0.02716928005218506
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,2,32,power_law_1.2,0.02778624057769775
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,2,32,power_law_1.2,0.04066048145294189
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,2,32,power_law_1.2,0.042885122299194334
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,2,32,power_law_1.2,0.04193024158477783
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,2,32,power_law_1.2,0.04818943977355957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,2,32,power_law_1.2,0.04581376075744629
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,2,32,power_law_1.2,0.04797311782836914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,2,32,power_law_1.2,0.05008128166198731
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,2,32,power_law_1.2,0.05635583877563477
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,2,32,power_law_1.2,0.05544960021972656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,2,32,power_law_1.2,0.054789118766784675
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,2,32,power_law_1.2,0.05339903831481933
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,2,32,power_law_1.2,0.057379841804504395
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,2,32,power_law_1.2,0.06167808055877686
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,2,32,power_law_1.2,0.06484864234924317
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,2,32,power_law_1.2,0.08538368225097656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,2,32,power_law_1.2,0.09084159851074218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,2,32,power_law_1.2,1.055477752685547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,2,32,power_law_1.2,0.1263526439666748
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,2,32,power_law_1.2,0.17745792388916015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,2,32,power_law_1.2,0.835203857421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,2,32,power_law_1.2,0.2559667205810547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,2,32,power_law_1.2,0.3335103988647461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,2,32,power_law_1.2,0.6353510284423829
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,4,16,balanced,0.02184959888458252
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,4,16,balanced,0.021862399578094483
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,4,16,balanced,0.022992639541625975
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,4,16,balanced,0.035704319477081296
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,1,64,power_law_1.2,3.650819702148438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,2,32,power_law_1.2,0.5305753707885742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,4,16,balanced,0.04576640129089356
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,4,16,balanced,0.0746995210647583
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,4,16,balanced,0.07962624073028565
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,4,16,balanced,0.07951488018035888
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,4,16,balanced,0.0805951976776123
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,4,16,balanced,0.0831935977935791
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,4,16,balanced,0.08702079772949219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,4,16,balanced,0.09075839996337891
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,4,16,balanced,0.08315520286560059
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,4,16,balanced,0.08522879600524902
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,4,16,balanced,0.08675456047058105
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,4,16,balanced,0.09044480323791504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,4,16,balanced,0.09806719779968262
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,4,16,balanced,0.0746342420578003
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,4,16,balanced,0.10524415969848633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,4,16,balanced,0.12311936378479005
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,4,16,balanced,0.13697664260864256
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,4,16,balanced,0.17120384216308593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,4,16,balanced,0.20136703491210936
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,4,16,balanced,0.2952230453491211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,4,16,balanced,0.37728897094726566
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,4,16,balanced,0.024033279418945314
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,4,16,balanced,0.5466342544555663
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,4,16,balanced,0.02388607978820801
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,4,16,balanced,0.02451200008392334
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,4,16,balanced,0.7074150085449219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,4,16,balanced,0.02470144033432007
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,4,16,balanced,0.028424959182739257
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,4,16,balanced,0.04562943935394287
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,4,16,balanced,0.046393599510192875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,4,16,balanced,0.051248641014099115
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,4,16,balanced,0.049180159568786616
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,4,16,balanced,0.05111680030822754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,4,16,balanced,0.05422080039978028
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,4,16,balanced,0.056363520622253416
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,4,16,balanced,0.06005760192871094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,1,64,power_law_1.2,1.702398681640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,4,16,balanced,0.051381759643554695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,4,16,balanced,0.05384191989898681
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,4,16,balanced,0.05506048202514648
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,4,16,balanced,0.06351232051849365
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,4,16,balanced,0.058685441017150876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,4,16,balanced,0.07022079944610596
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,4,16,balanced,0.08948479652404785
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,4,16,balanced,0.1038105583190918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,4,16,balanced,0.1367155170440674
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,4,16,balanced,0.16572160720825196
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,4,16,balanced,0.22775680541992188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,4,16,balanced,0.28609920501708985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,4,16,power_law_1.01,0.04475135803222656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,4,16,balanced,0.4293875122070313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,4,16,power_law_1.01,0.04482944011688232
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,4,16,power_law_1.01,0.045029120445251467
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,4,16,power_law_1.01,0.045154561996459966
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,4,16,power_law_1.01,0.07060863971710205
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,4,16,power_law_1.01,0.07274752140045167
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,4,16,power_law_1.01,0.07575295925140381
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,4,16,power_law_1.01,0.07763840198516846
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,4,16,power_law_1.01,0.0806873607635498
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,4,16,power_law_1.01,0.08388863563537598
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,4,16,power_law_1.01,0.0868620777130127
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,4,16,power_law_1.01,0.087511043548584
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,4,16,power_law_1.01,0.08608384132385254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,4,16,power_law_1.01,0.08790528297424316
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,4,16,power_law_1.01,0.09058048248291015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,4,16,power_law_1.01,0.0932646369934082
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,4,16,balanced,0.5498086547851562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,4,16,power_law_1.01,0.11233663558959961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,4,16,power_law_1.01,0.12535679817199707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,4,16,power_law_1.01,0.1650649642944336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,4,16,power_law_1.01,0.1899443244934082
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,4,16,power_law_1.01,0.24271104812622069
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,4,16,power_law_1.01,0.3451916885375977
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,4,16,power_law_1.01,0.492968978881836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,4,16,power_law_1.01,0.09844863891601563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,4,16,power_law_1.01,0.6512525177001953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,4,16,power_law_1.01,0.025783040523529054
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,4,16,power_law_1.01,0.029713919162750246
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,4,16,power_law_1.01,0.927188491821289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,4,16,power_law_1.01,0.02933631896972656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,4,16,power_law_1.01,0.04699903964996338
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,4,16,power_law_1.01,0.04192768096923828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,4,16,power_law_1.01,1.2523852539062499
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,4,16,power_law_1.01,0.04907135963439942
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,4,16,power_law_1.01,0.051215357780456536
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,4,16,power_law_1.01,0.05225728034973145
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,4,16,power_law_1.01,0.029374721050262453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,4,16,power_law_1.01,0.057043199539184575
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,4,16,power_law_1.01,0.05844863891601563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,4,16,power_law_1.01,0.057504000663757326
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,4,16,power_law_1.01,0.05529727935791016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,4,16,power_law_1.01,0.05728640079498291
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,4,16,power_law_1.01,0.059516158103942875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,4,16,power_law_1.01,0.06382336139678954
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,4,16,power_law_1.01,0.07655424118041992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,2,32,power_law_1.2,1.536180419921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,4,16,power_law_1.01,0.09325311660766603
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,4,16,power_law_1.01,0.048830718994140626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,4,16,power_law_1.01,0.12373632431030272
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,4,16,power_law_1.01,0.151147518157959
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,4,16,power_law_1.01,0.16303232192993164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,4,16,power_law_1.01,0.2789324760437012
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,4,16,power_law_1.01,0.362872314453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,4,16,power_law_1.01,0.4899929428100586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,4,16,power_law_1.2,0.044592638015747074
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,4,16,power_law_1.2,0.04414976119995117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,4,16,power_law_1.2,0.044939517974853516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,4,16,power_law_1.01,0.8067327880859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,4,16,power_law_1.2,0.04468992233276367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,4,16,power_law_1.2,0.06759168148040771
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,4,16,power_law_1.2,0.07395328044891357
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,4,16,power_law_1.2,0.07752319812774658
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,4,16,power_law_1.2,0.07878784179687501
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,4,16,power_law_1.2,0.07971199989318847
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,4,16,power_law_1.2,0.08269696235656739
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,4,16,power_law_1.2,0.08875136375427246
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,4,16,power_law_1.2,0.09045120239257812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,4,16,power_law_1.2,0.08425215721130372
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,4,16,power_law_1.2,0.08706432342529297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,4,16,power_law_1.2,0.0908198356628418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,4,16,power_law_1.2,0.09443455696105957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,4,16,power_law_1.2,0.10098943710327149
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,2,32,power_law_1.2,0.9664959716796876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,4,16,power_law_1.2,0.11963264465332031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,4,16,power_law_1.2,0.14232959747314453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,4,16,power_law_1.2,0.19808128356933594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,4,16,power_law_1.2,0.22305408477783203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,4,16,power_law_1.2,0.4277030563354492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,4,16,power_law_1.01,0.9539225769042968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,4,16,power_law_1.2,0.026211841106414797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,4,16,power_law_1.2,0.34528385162353514
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,4,16,power_law_1.2,0.02904831886291504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,4,16,power_law_1.2,0.029510400295257568
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,4,16,power_law_1.2,0.029273600578308107
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,4,16,power_law_1.2,0.04294271945953369
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,4,16,power_law_1.2,0.04536575794219971
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,2,32,power_law_1.2,1.2331648254394532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,4,16,power_law_1.2,0.04804992198944092
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,4,16,power_law_1.2,0.04875391960144043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,4,16,power_law_1.2,0.053754878044128415
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,4,16,power_law_1.2,0.059214081764221195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,4,16,power_law_1.2,0.05751423835754395
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,4,16,power_law_1.2,0.05547904014587403
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,4,16,power_law_1.2,0.05729152202606201
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,4,16,power_law_1.2,0.05874176025390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,4,16,power_law_1.2,0.06107903957366943
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,4,16,power_law_1.2,0.06629119873046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,4,16,power_law_1.2,0.08610688209533691
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,4,16,power_law_1.2,0.09567999839782715
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,4,16,power_law_1.2,0.05209599971771241
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,4,16,power_law_1.2,0.11647744178771972
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,4,16,power_law_1.2,0.1616307258605957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,4,16,power_law_1.2,0.28848640441894535
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,4,16,power_law_1.2,0.6280588912963867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,4,16,power_law_1.2,0.8261824035644532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,4,16,power_law_1.2,0.2350387191772461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,8,8,balanced,0.021867520809173584
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,8,8,balanced,0.022709760665893555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,8,8,balanced,0.024206080436706544
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,8,8,balanced,0.03732352018356323
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,8,8,balanced,0.04853888034820557
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,8,8,balanced,0.07542272090911865
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,8,8,balanced,0.07989376068115235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,8,8,balanced,0.0832908821105957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,8,8,balanced,0.08382847785949707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,2,32,power_law_1.2,2.2189733886718748
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,4,16,power_law_1.2,0.43770111083984375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,8,8,balanced,0.08806143760681152
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,8,8,balanced,0.08485631942749024
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,8,8,balanced,0.08881535530090331
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,8,8,balanced,0.09181695938110351
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,8,8,balanced,0.09523072242736816
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,8,8,balanced,0.10223872184753417
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,8,8,balanced,0.11689855575561522
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,8,8,balanced,0.13070464134216309
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,8,8,balanced,0.08240511894226074
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,8,8,balanced,0.16200704574584962
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,8,8,balanced,0.18890623092651365
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,8,8,balanced,0.24628095626831054
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,8,8,balanced,0.3049113655090332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,8,8,balanced,0.4471334457397461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,8,8,balanced,0.08398719787597655
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,8,8,balanced,0.589062385559082
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,8,8,balanced,0.024296960830688476
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,8,8,balanced,0.024647679328918457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,8,8,balanced,0.8476518249511719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,8,8,balanced,0.02694272041320801
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,8,8,balanced,1.117822723388672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,8,8,balanced,0.034149119853973384
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,8,8,balanced,0.05419904232025147
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,8,8,balanced,0.05315328121185303
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,8,8,balanced,0.05782271862030029
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,8,8,balanced,0.06133632183074951
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,8,8,balanced,0.06300415992736816
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,8,8,balanced,0.05560704231262207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,8,8,balanced,0.058097920417785644
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,8,8,balanced,0.0247871994972229
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,8,8,balanced,0.05945727825164795
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,8,8,balanced,0.062431998252868645
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,8,8,balanced,0.06354688167572022
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,8,8,balanced,0.0656115198135376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,8,8,balanced,0.07319424152374268
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,8,8,balanced,0.08934528350830077
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,8,8,balanced,0.10371968269348145
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,4,16,power_law_1.2,0.5915135955810547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,8,8,balanced,0.1313024044036865
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,8,8,balanced,0.15830528259277343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,8,8,balanced,0.22126848220825196
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,8,8,balanced,0.2727475166320801
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,8,8,balanced,0.38887935638427734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,8,8,balanced,0.5031961441040039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,8,8,power_law_1.01,0.03458559989929199
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,8,8,balanced,0.7529420471191406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,8,8,power_law_1.01,0.04131840229034424
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,8,8,power_law_1.01,0.04316544055938721
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,8,8,balanced,0.9800678253173828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,8,8,power_law_1.01,0.04587520122528076
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,8,8,power_law_1.01,0.046813440322875974
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,8,8,power_law_1.01,0.07552383899688721
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,8,8,power_law_1.01,0.07349631786346436
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,8,8,power_law_1.01,0.07924352169036866
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,8,8,power_law_1.01,0.08275327682495118
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,8,8,power_law_1.01,0.08510720252990722
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,8,8,power_law_1.01,0.08781184196472168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,8,8,power_law_1.01,0.08621952056884766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,8,8,power_law_1.01,0.0866380786895752
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,8,8,power_law_1.01,0.08943488121032714
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,8,8,power_law_1.01,0.08676351547241211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,8,8,power_law_1.01,0.09285120010375977
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,8,8,power_law_1.01,0.09971327781677246
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,8,8,power_law_1.01,0.11627136230468751
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,8,8,power_law_1.01,0.12703231811523436
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,8,8,power_law_1.01,0.16048383712768555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,8,8,power_law_1.01,0.19734783172607423
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,8,8,power_law_1.01,0.2608716773986816
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,8,8,power_law_1.01,0.33128704071044923
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,8,8,power_law_1.01,0.4524249649047851
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,8,8,power_law_1.01,0.6424089813232422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,8,8,power_law_1.01,0.02598016023635864
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,8,8,power_law_1.01,0.03162623882293701
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,8,8,power_law_1.01,0.03242239952087402
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,8,8,power_law_1.01,0.9060160064697266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,8,8,power_law_1.01,0.033413119316101074
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,8,8,power_law_1.01,0.034296319484710694
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,8,8,power_law_1.01,0.04932479858398438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,8,8,power_law_1.01,1.180337905883789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,8,8,power_law_1.01,0.05291520118713379
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,8,8,power_law_1.01,0.05454976081848144
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,8,8,power_law_1.01,0.057079038619995116
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,8,8,power_law_1.01,0.05923200130462647
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,8,8,power_law_1.01,0.06380544185638429
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,8,8,power_law_1.01,0.061352958679199224
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,8,8,power_law_1.01,0.05933440208435059
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,4,16,power_law_1.2,1.2145740509033203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,8,8,power_law_1.01,0.05971199989318847
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,8,8,power_law_1.01,0.061684479713439946
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,8,8,power_law_1.01,0.06377344131469727
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,8,8,power_law_1.01,0.06835072040557862
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,8,8,power_law_1.01,0.09931391716003418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,8,8,power_law_1.01,0.13530367851257324
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,8,8,power_law_1.01,0.16005376815795896
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,8,8,power_law_1.01,0.21671936035156253
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,8,8,power_law_1.01,0.28167295455932617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,8,8,power_law_1.01,0.4220620727539062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,8,8,power_law_1.01,0.5107686233520508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,8,8,power_law_1.2,0.024796159267425538
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,8,8,power_law_1.2,0.039406080245971684
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,8,8,power_law_1.01,0.08676608085632324
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,8,8,power_law_1.01,0.8013529968261718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,8,8,power_law_1.2,0.04174592018127442
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,8,8,power_law_1.2,0.04676352024078369
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,8,8,power_law_1.2,0.0449241590499878
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,8,8,power_law_1.2,0.07397632122039795
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,8,8,power_law_1.2,0.07995776176452637
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,8,8,power_law_1.2,0.08178303718566896
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,8,8,power_law_1.2,0.0835801601409912
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,4,16,power_law_1.2,0.8547007751464843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,8,8,power_law_1.2,0.09131008148193359
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,8,8,power_law_1.2,0.08710271835327148
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,8,8,power_law_1.2,0.08496383666992188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,8,8,power_law_1.2,0.08926464080810546
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,8,8,power_law_1.2,0.09185664176940918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,8,8,power_law_1.2,0.09632639884948731
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,8,8,power_law_1.2,0.10173952102661132
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,8,8,power_law_1.2,0.11814656257629394
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,8,8,power_law_1.2,0.13399807929992674
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,8,8,power_law_1.2,0.1682086372375488
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,8,8,power_law_1.2,0.2044927978515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,8,8,power_law_1.2,0.07726463794708252
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,8,8,power_law_1.01,0.9962163543701171
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,8,8,power_law_1.2,0.3909465789794922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,8,8,power_law_1.2,0.3033497619628906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,8,8,power_law_1.2,0.026051840782165527
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,8,8,power_law_1.2,0.027504639625549315
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,8,8,power_law_1.2,0.03131392002105713
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,8,8,power_law_1.2,0.03377279996871948
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,8,8,power_law_1.2,0.03391232013702393
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,8,8,power_law_1.2,0.05219327926635742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,8,8,power_law_1.2,0.05276288032531738
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,8,8,power_law_1.2,0.05562111854553222
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,8,8,power_law_1.2,0.057631998062133795
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,8,8,power_law_1.2,0.05923967838287354
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,8,8,power_law_1.2,0.06025343894958496
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,8,8,power_law_1.2,0.05677696228027344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,8,8,power_law_1.2,0.058311681747436526
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,8,8,power_law_1.2,0.7175065612792968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,8,8,power_law_1.2,0.06263167858123779
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,8,8,power_law_1.2,0.06358016014099122
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,8,8,power_law_1.2,0.0662604808807373
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,8,8,power_law_1.2,0.0728332805633545
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,8,8,power_law_1.2,0.5549619293212891
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,8,8,power_law_1.2,0.08909312248229981
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,8,8,power_law_1.2,0.10642815589904786
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,8,8,power_law_1.2,0.1374630355834961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,8,8,power_law_1.2,0.16680448532104492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,8,8,power_law_1.2,0.23060863494873046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,8,8,power_law_1.2,0.3074995231628418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,4,16,power_law_1.2,1.6286553955078127
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,16,4,balanced,0.023119359016418456
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,16,4,balanced,0.024583680629730226
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,8,8,power_law_1.2,0.5551142501831055
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,16,4,balanced,0.030150399208068845
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,16,4,balanced,0.0410752010345459
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,8,8,power_law_1.2,0.43221374511718746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,16,4,balanced,0.05465472221374512
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,16,4,balanced,0.056789760589599614
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,16,4,balanced,0.08281087875366211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,16,4,balanced,0.08563455581665039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,16,4,balanced,0.0884160041809082
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,16,4,balanced,0.0911564826965332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,16,4,balanced,0.08219136238098144
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,16,4,balanced,0.08465920448303223
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,16,4,balanced,0.08733440399169921
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,16,4,balanced,0.08825087547302246
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,16,4,balanced,0.09230079650878906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,16,4,balanced,0.09573504447937012
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,16,4,balanced,0.10118144035339356
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,16,4,balanced,0.11652223587036133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,16,4,balanced,0.12988800048828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,16,4,balanced,0.16185216903686522
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,16,4,balanced,0.18802688598632813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,16,4,balanced,0.24488447189331053
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,16,4,balanced,0.3017036819458008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,16,4,balanced,0.43856895446777344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,16,4,balanced,0.5827980804443359
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,16,4,balanced,0.02482815980911255
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,16,4,balanced,0.8557952117919921
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,16,4,balanced,0.025429759025573727
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,16,4,balanced,0.027878398895263674
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,16,4,balanced,0.031374080181121825
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,16,4,balanced,0.04385024070739746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,16,4,balanced,1.130613784790039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,16,4,balanced,0.04715007781982422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,16,4,balanced,0.06296063899993896
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,8,8,power_law_1.2,1.0174745941162109
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,16,4,balanced,0.06587264060974121
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,16,4,balanced,0.07062016010284425
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,16,4,balanced,0.06521471977233886
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,16,4,balanced,0.06662271976470947
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,16,4,balanced,0.06946688175201417
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,4,16,power_law_1.2,1.1360422515869142
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,16,4,balanced,0.07021183967590332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,16,4,balanced,0.07296383857727051
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,16,4,balanced,0.07909759998321533
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,16,4,balanced,0.09530240058898926
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,16,4,balanced,0.10778112411499023
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,16,4,balanced,0.1350822353363037
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,16,4,balanced,0.15881983757019041
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,16,4,balanced,0.22759679794311521
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,16,4,balanced,0.06587776184082031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,16,4,balanced,0.27552127838134766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,16,4,balanced,0.06232831954956055
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,16,4,balanced,0.3923507308959961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,16,4,power_law_1.01,0.024961280822753906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,16,4,balanced,0.5103513717651367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,16,4,power_law_1.01,0.03899904012680054
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,16,4,power_law_1.01,0.04123648166656494
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,16,4,power_law_1.01,0.04446080207824707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,16,4,balanced,0.7671974182128907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,16,4,power_law_1.01,0.08074496269226075
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,16,4,power_law_1.01,0.0841100788116455
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,16,4,balanced,1.0006873321533203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,16,4,power_law_1.01,0.08565119743347169
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,16,4,power_law_1.01,0.08899840354919433
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,16,4,power_law_1.01,0.09059455871582031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,16,4,power_law_1.01,0.08549375534057617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,16,4,power_law_1.01,0.08933631896972656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,16,4,power_law_1.01,0.09188480377197265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,16,4,power_law_1.01,0.09568896293640136
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,16,4,power_law_1.01,0.10300671577453613
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,16,4,power_law_1.01,0.11847295761108398
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,16,4,power_law_1.01,0.1322214412689209
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,16,4,power_law_1.01,0.16333183288574218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,16,4,power_law_1.01,0.08357888221740722
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,16,4,power_law_1.01,0.049610238075256344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,16,4,power_law_1.01,0.054214401245117186
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,16,4,power_law_1.01,0.25615615844726564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,16,4,power_law_1.01,0.31568384170532227
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,16,4,power_law_1.01,0.47843070983886715
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,16,4,power_law_1.01,0.6220492935180664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,16,4,power_law_1.01,0.026182401180267333
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,16,4,power_law_1.01,0.19538175582885742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,16,4,power_law_1.01,0.9207027435302735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,16,4,power_law_1.01,0.029447679519653318
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,16,4,power_law_1.01,0.031544320583343506
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,16,4,power_law_1.01,0.03537408113479614
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,16,4,power_law_1.01,0.038475520610809326
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,16,4,power_law_1.01,0.04452991962432861
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,16,4,power_law_1.01,0.060478720664978024
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,16,4,power_law_1.01,0.06424320220947266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,16,4,power_law_1.01,1.2202444458007813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,16,4,power_law_1.01,0.06567168235778809
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,16,4,power_law_1.01,0.06790527820587158
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,16,4,power_law_1.01,0.06662528038024902
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,16,4,power_law_1.01,0.06290559768676758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,16,4,power_law_1.01,0.0653875207901001
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,16,4,power_law_1.01,0.06836863994598388
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,16,4,power_law_1.01,0.0691315221786499
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,16,4,power_law_1.01,0.07166848182678223
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,16,4,power_law_1.01,0.07864704132080078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,16,4,power_law_1.01,0.09640192031860352
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,16,4,power_law_1.01,0.11258111953735353
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,16,4,power_law_1.01,0.14288640022277832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,16,4,power_law_1.01,0.1761075210571289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,16,4,power_law_1.01,0.23341440200805663
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,16,4,power_law_1.01,0.3037491226196289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,16,4,power_law_1.01,0.44436096191406255
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,16,4,power_law_1.2,0.024902400970458986
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,16,4,power_law_1.01,0.5686899185180664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,16,4,power_law_1.2,0.03784192085266113
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,16,4,power_law_1.2,0.04133887767791748
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,16,4,power_law_1.2,0.0449241590499878
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,16,4,power_law_1.2,0.049120001792907715
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,16,4,power_law_1.2,0.05502975940704345
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,16,4,power_law_1.2,0.08123392105102539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,16,4,power_law_1.2,0.0841420841217041
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,16,4,power_law_1.01,1.0797017669677733
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,16,4,power_law_1.2,0.0858739185333252
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,16,4,power_law_1.2,0.0884556770324707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,16,4,power_law_1.2,0.08949631690979004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,16,4,power_law_1.2,0.08461440086364747
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,16,4,power_law_1.2,0.08724224090576171
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,16,4,power_law_1.2,0.09098752021789551
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,16,4,power_law_1.2,0.09287039756774902
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,16,4,power_law_1.2,0.09725055694580079
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,16,4,power_law_1.2,0.1042137622833252
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,16,4,power_law_1.2,0.12096384048461914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,16,4,power_law_1.2,0.13646335601806642
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,16,4,power_law_1.2,0.1673971176147461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,16,4,power_law_1.2,0.1963711929321289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,16,4,power_law_1.01,0.8413030242919921
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,16,4,power_law_1.2,0.3421299362182617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,8,8,power_law_1.2,0.7753215789794922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,16,4,power_law_1.2,0.26918783187866213
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,16,4,power_law_1.2,0.026206719875335693
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,16,4,power_law_1.2,0.02756736040115356
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,16,4,power_law_1.2,0.030553600788116454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,16,4,power_law_1.2,0.035272960662841794
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,16,4,power_law_1.2,0.03758080005645752
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,16,4,power_law_1.2,0.04457727909088135
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,16,4,power_law_1.2,0.06077695846557617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,16,4,power_law_1.2,0.06428160190582274
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,16,4,power_law_1.2,0.06639232158660888
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,16,4,power_law_1.2,0.06898560047149657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,8,8,power_law_1.2,1.3110617065429688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,16,4,power_law_1.2,0.062431998252868645
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,16,4,power_law_1.2,0.06470143795013428
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,16,4,power_law_1.2,0.06601088047027588
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,16,4,power_law_1.2,0.07059199810028076
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,16,4,power_law_1.2,0.07337471961975098
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,16,4,power_law_1.2,0.08162176132202148
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,16,4,power_law_1.2,0.09944576263427735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,16,4,power_law_1.2,0.11482879638671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,16,4,power_law_1.2,0.6454643249511719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,16,4,power_law_1.2,0.14308095932006837
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,16,4,power_law_1.2,0.17714176177978516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,16,4,power_law_1.2,0.0694323205947876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,16,4,power_law_1.2,0.31445119857788084
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,16,4,power_law_1.2,0.24257024765014648
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,16,4,power_law_1.2,0.49500545501708987
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,32,2,balanced,0.022613759040832522
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,32,2,balanced,0.024430079460144045
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,32,2,balanced,0.029012479782104493
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,32,2,balanced,0.04462080001831055
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,32,2,balanced,0.05756800174713135
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,32,2,balanced,0.05921792030334473
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,32,2,balanced,0.06098432064056396
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,32,2,balanced,0.06250495910644531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,32,2,balanced,0.08362624168395996
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,32,2,balanced,0.0844595241546631
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,32,2,balanced,0.08594304084777832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,16,4,power_law_1.2,0.45012992858886725
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,32,2,balanced,0.08909440040588379
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,8,8,power_law_1.2,1.0805542755126953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,32,2,balanced,0.09085311889648437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,32,2,balanced,0.09332863807678222
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,32,2,balanced,0.09500415802001953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,32,2,balanced,0.10322048187255858
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,32,2,balanced,0.11661567687988281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,32,2,balanced,0.13130111694335939
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,32,2,balanced,0.1654323196411133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,32,2,balanced,0.1910361671447754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,32,2,balanced,0.09673600196838379
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,32,2,balanced,0.24938751220703126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,32,2,balanced,0.3088256072998047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,32,2,balanced,0.48450046539306635
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,32,2,balanced,0.6233497619628906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,32,2,balanced,0.9343705749511718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,32,2,balanced,0.024312319755554198
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,32,2,balanced,0.025759999752044675
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,32,2,balanced,0.02948992013931274
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,32,2,balanced,0.037304320335388184
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,32,2,balanced,0.053509120941162114
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,32,2,balanced,0.055530238151550296
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,32,2,balanced,0.05521791934967041
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,16,4,power_law_1.2,0.585747184753418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,32,2,balanced,1.243246078491211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,32,2,balanced,0.05646975994110107
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,32,2,balanced,0.07847296237945557
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,32,2,balanced,0.07937535762786865
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,32,2,balanced,0.08085375785827636
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,32,2,balanced,0.08322303771972657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,32,2,balanced,0.08445055961608887
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,32,2,balanced,0.08780672073364257
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,32,2,balanced,0.08874367713928223
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,32,2,balanced,0.09618687629699707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,32,2,balanced,0.10954239845275879
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,32,2,balanced,0.12274815559387206
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,32,2,balanced,0.14941439628601075
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,32,2,balanced,0.17340032577514647
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,32,2,balanced,0.2544384002685547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,32,2,balanced,0.30103679656982424
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,32,2,balanced,0.4515647888183594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,32,2,balanced,0.09034879684448242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,32,2,balanced,0.5829119873046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,32,2,power_law_1.01,0.02360192060470581
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,32,2,power_law_1.01,0.026521599292755126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,32,2,balanced,0.86487548828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,32,2,power_law_1.01,0.04069632053375244
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,32,2,power_law_1.01,0.0452672004699707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,32,2,power_law_1.01,0.04828415870666504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,32,2,power_law_1.01,0.056139521598815925
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,32,2,power_law_1.01,0.05859712123870849
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,32,2,balanced,1.1467788696289063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,32,2,power_law_1.01,0.06048511981964111
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,32,2,power_law_1.01,0.0824512004852295
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,32,2,power_law_1.01,0.08389760017395019
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,32,2,power_law_1.01,0.08754303932189941
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,32,2,power_law_1.01,0.08560511589050293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,32,2,power_law_1.01,0.09432064056396484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,32,2,power_law_1.01,0.09664511680603027
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,32,2,power_law_1.01,0.09950592041015624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,32,2,power_law_1.01,0.10724736213684083
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,32,2,power_law_1.01,0.12259712219238281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,32,2,power_law_1.01,0.13828736305236816
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,32,2,power_law_1.01,0.1699929618835449
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,32,2,power_law_1.01,0.20168447494506836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,32,2,power_law_1.01,0.2632486343383789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,32,2,power_law_1.01,0.3541491317749023
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,32,2,power_law_1.01,0.08939264297485353
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,32,2,power_law_1.01,0.5168012619018555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,32,2,power_law_1.01,0.6869503784179687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,32,2,power_law_1.01,0.025498878955841065
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,32,2,power_law_1.01,0.03276416063308716
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,32,2,power_law_1.01,0.9903155517578124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,32,2,power_law_1.01,0.03902591943740845
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,32,2,power_law_1.01,0.04158463954925537
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,32,2,power_law_1.01,1.2809036254882813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,32,2,power_law_1.01,0.05208447933197021
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,32,2,power_law_1.01,0.05255424022674561
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,32,2,power_law_1.01,0.0745907211303711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,32,2,power_law_1.01,0.0763481616973877
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,32,2,power_law_1.01,0.07890304088592529
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,16,4,power_law_1.2,0.9786431884765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,32,2,power_law_1.01,0.028523519039154056
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,32,2,power_law_1.01,0.08218751907348633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,32,2,power_law_1.01,0.08359423637390137
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,32,2,power_law_1.01,0.08620287895202636
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,32,2,power_law_1.01,0.08738559722900391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,32,2,power_law_1.01,0.09072383880615234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,32,2,power_law_1.01,0.09882752418518066
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,32,2,power_law_1.01,0.11679615974426269
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,32,2,power_law_1.01,0.13387007713317872
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,32,2,power_law_1.01,0.17034879684448243
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,32,2,power_law_1.01,0.20450944900512696
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,32,2,power_law_1.01,0.05375872135162354
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,32,2,power_law_1.01,0.2919424057006836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,32,2,power_law_1.01,0.363265266418457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,32,2,power_law_1.01,0.5102860641479492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,32,2,power_law_1.2,0.02341376066207886
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,32,2,power_law_1.01,0.654207992553711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,32,2,power_law_1.2,0.040318717956542974
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,32,2,power_law_1.2,0.04465792179107666
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,32,2,power_law_1.01,0.9857907104492188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,32,2,power_law_1.2,0.0562227201461792
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,32,2,power_law_1.2,0.05802239894866943
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,32,2,power_law_1.01,1.2346316528320311
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,32,2,power_law_1.2,0.06070015907287598
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,32,2,power_law_1.2,0.08127615928649903
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,32,2,power_law_1.2,0.08578304290771485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,32,2,power_law_1.2,0.030205440521240235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,32,2,power_law_1.2,0.08820608139038086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,32,2,power_law_1.2,0.09085311889648437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,32,2,power_law_1.2,0.04739327907562256
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,32,2,power_law_1.2,0.0950169563293457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,32,2,power_law_1.2,0.09699071884155273
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,32,2,power_law_1.2,0.10056832313537598
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,32,2,power_law_1.2,0.10699520111083985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,32,2,power_law_1.2,0.12326144218444823
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,32,2,power_law_1.2,0.13803775787353517
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,32,2,power_law_1.2,0.1709427261352539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,32,2,power_law_1.2,0.203558406829834
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,16,4,power_law_1.2,0.8515110778808594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,32,2,power_law_1.2,0.3648384094238281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,32,2,power_law_1.2,0.2647577667236328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,32,2,power_law_1.2,0.08311552047729492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,32,2,power_law_1.2,0.025280001163482668
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,32,2,power_law_1.2,0.02818687915802002
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,32,2,power_law_1.2,0.0314086389541626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,32,2,power_law_1.2,0.037039361000061034
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,32,2,power_law_1.2,0.04202623844146729
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,32,2,power_law_1.2,0.051868162155151366
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,32,2,power_law_1.2,0.0512332820892334
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,32,2,power_law_1.2,0.05464704036712646
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,32,2,power_law_1.2,0.07602943897247313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,32,2,power_law_1.2,0.07640448093414307
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,32,2,power_law_1.2,0.07913343906402588
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,32,2,power_law_1.2,0.08330880165100098
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,32,2,power_law_1.2,0.08405119895935058
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,32,2,power_law_1.2,0.08794879913330078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,32,2,power_law_1.2,0.08889984130859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,32,2,power_law_1.2,0.0921996784210205
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,32,2,power_law_1.2,0.09922047615051269
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,32,2,power_law_1.2,0.11894783973693848
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,32,2,power_law_1.2,0.13618687629699705
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,32,2,power_law_1.2,0.17138944625854494
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,32,2,power_law_1.2,0.20676864624023436
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,32,2,power_law_1.2,0.29441152572631835
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,32,2,power_law_1.2,0.36673152923583985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,32,2,power_law_1.2,0.5358579254150391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,16,4,power_law_1.2,1.2567769622802734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,32,2,power_law_1.2,0.6977958679199219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,32,2,power_law_1.2,0.5130316925048828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,2,32,balanced,0.05351168155670166
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,2,32,balanced,0.05305088043212891
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,2,32,balanced,0.05357439994812012
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,2,32,balanced,0.06011263847351074
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,2,32,balanced,0.07399807929992676
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,32,2,power_law_1.2,0.9990528106689454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,2,32,balanced,0.129932804107666
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,2,32,balanced,0.13026816368103028
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,2,32,balanced,0.13091327667236327
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,2,32,balanced,0.13352191925048829
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,2,32,balanced,0.13373696327209472
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,2,32,balanced,0.1351244831085205
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,2,32,balanced,0.13948415756225585
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,2,32,balanced,0.1426585578918457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,2,32,balanced,0.14466048240661622
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,2,32,balanced,0.14849791526794434
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,2,32,balanced,0.14563584327697754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,2,32,balanced,0.15568511962890624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,2,32,balanced,0.16391807556152344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,2,32,balanced,0.1308454418182373
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,2,32,balanced,0.18411008834838866
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,2,32,balanced,0.2009459114074707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,2,32,balanced,0.27497215270996095
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,2,32,balanced,0.33947521209716797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,2,32,balanced,0.43566207885742186
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,2,32,balanced,0.5768384170532227
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,2,32,balanced,0.030595839023590088
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,2,32,balanced,0.790540771484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,2,32,balanced,0.03096191883087158
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,2,32,balanced,0.030700800418853758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,2,32,balanced,0.03738111972808838
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,2,32,balanced,1.0298995208740234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,2,32,balanced,0.04787839889526367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,2,32,balanced,0.06311423778533935
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,2,32,balanced,0.0646937608718872
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,32,2,power_law_1.2,0.6662770843505859
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,2,32,balanced,0.06705535888671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,2,32,balanced,0.06759679794311524
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,2,32,balanced,0.06903423786163329
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,2,32,balanced,0.07205503940582275
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,2,32,balanced,0.07505663871765136
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,2,32,balanced,0.07785600185394287
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,2,32,balanced,0.08105216026306153
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,2,32,balanced,0.07647488117218018
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,2,32,balanced,0.08306431770324707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,2,32,balanced,0.0872435188293457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,2,32,balanced,0.06523392200469971
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,2,32,balanced,0.0997824001312256
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,2,32,balanced,0.11157376289367675
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,2,32,balanced,0.1445094394683838
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,2,32,balanced,0.16788223266601562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,2,32,balanced,0.24767488479614258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,2,32,balanced,0.29861375808715823
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,2,32,balanced,0.3980940628051758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,2,32,power_law_1.01,0.08165632247924805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,2,32,balanced,0.5318131256103515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,2,32,power_law_1.01,0.08108672142028808
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,2,32,balanced,0.06568448066711426
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,2,32,power_law_1.01,0.08133376121520997
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,2,32,power_law_1.01,0.12702207565307616
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,2,32,power_law_1.01,0.11944704055786133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,2,32,power_law_1.01,0.12669568061828612
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,2,32,power_law_1.01,0.12759807586669922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,32,2,power_law_1.2,0.9597388458251952
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,2,32,power_law_1.01,0.12707967758178712
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,2,32,power_law_1.01,0.1290675163269043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,2,32,power_law_1.01,0.12883456230163576
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,2,32,power_law_1.01,0.13967103958129884
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,2,32,power_law_1.01,0.14088447570800783
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,2,32,power_law_1.01,0.14595711708068848
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,2,32,power_law_1.01,0.14994048118591308
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,2,32,power_law_1.01,0.15095935821533205
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,2,32,power_law_1.01,0.15978879928588868
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,2,32,power_law_1.01,0.17168895721435545
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,2,32,power_law_1.01,0.20466560363769531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,2,32,power_law_1.01,0.2570841598510742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,2,32,power_law_1.01,0.3275455856323242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,2,32,power_law_1.01,0.40160255432128905
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,2,32,power_law_1.01,0.5738304138183594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,2,32,power_law_1.01,0.746792984008789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,2,32,power_law_1.01,0.9832243347167969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,2,32,power_law_1.01,0.05102591991424561
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,2,32,power_law_1.01,0.051167998313903805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,2,32,power_law_1.01,0.0516044807434082
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,2,32,power_law_1.01,0.05942783832550049
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,2,32,power_law_1.01,1.6059609985351564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,2,32,power_law_1.01,0.059311361312866205
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,2,32,power_law_1.01,0.061423358917236326
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,2,32,power_law_1.01,0.06677375793457031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,2,32,power_law_1.01,0.0670143985748291
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,2,32,power_law_1.01,0.06969088077545166
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,2,32,power_law_1.01,0.07217408180236816
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,2,32,power_law_1.01,2.1135052490234374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,2,32,power_law_1.01,0.075665922164917
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,2,32,power_law_1.01,0.08251520156860351
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,2,32,power_law_1.01,0.08053119659423828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,2,32,power_law_1.01,0.08258175849914551
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,2,32,power_law_1.01,0.08482815742492675
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,2,32,power_law_1.01,0.09110015869140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,2,32,power_law_1.01,0.1090214443206787
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,2,32,power_law_1.01,0.06371327877044677
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,2,32,power_law_1.01,0.1529587173461914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,2,32,power_law_1.01,0.20885120391845705
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,2,32,power_law_1.01,3.06221435546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,2,32,power_law_1.01,0.2896396827697754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,2,32,power_law_1.01,0.373834228515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,2,32,power_law_1.01,0.5597439956665039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,2,32,power_law_1.01,0.1247424030303955
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,2,32,power_law_1.2,0.08163840293884277
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,2,32,power_law_1.01,0.7973900604248046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,2,32,power_law_1.2,0.08088319778442384
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,2,32,power_law_1.2,0.08166655540466308
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,2,32,power_law_1.01,1.1450726318359377
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,2,32,power_law_1.2,0.07337215900421143
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,2,32,power_law_1.2,0.12028800010681154
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,2,32,power_law_1.2,0.12766079902648925
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,2,32,power_law_1.2,0.12829695701599123
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,2,32,power_law_1.2,0.1288102436065674
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,2,32,power_law_1.01,1.3483084106445313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,2,32,power_law_1.2,0.13453056335449218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,2,32,power_law_1.2,0.14315135955810548
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,2,32,power_law_1.2,0.1495296001434326
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,2,32,power_law_1.2,0.1606118392944336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,2,32,power_law_1.2,0.17225727081298828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,2,32,power_law_1.2,0.18685184478759767
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,2,32,power_law_1.2,0.23346176147460937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,2,32,power_law_1.2,0.1283737564086914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,2,32,power_law_1.2,0.28748287200927736
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,2,32,power_law_1.2,0.14480768203735353
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,2,32,power_law_1.2,0.41269760131835936
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,2,32,power_law_1.2,0.4922150421142578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,2,32,power_law_1.2,0.8223423767089845
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,2,32,power_law_1.2,1.0987289428710938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,2,32,power_law_1.2,0.1512281608581543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,2,32,power_law_1.2,0.05165952205657959
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,2,32,power_law_1.2,1.5871334838867188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,16,4,power_law_1.2,1.090338592529297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,2,32,power_law_1.2,0.05089791774749756
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,2,32,power_law_1.2,0.0518400001525879
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,2,32,power_law_1.2,0.06066815853118897
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,2,32,power_law_1.2,0.0593779182434082
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,2,32,power_law_1.2,0.0613708782196045
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,2,32,power_law_1.2,0.06530432224273681
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,2,32,power_law_1.2,0.06723199844360352
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,2,32,power_law_1.2,0.07061888217926025
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,2,32,power_law_1.2,0.07240320205688476
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,2,32,power_law_1.2,0.08155903816223145
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,2,32,power_law_1.2,0.08086015701293944
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,2,32,power_law_1.2,0.08221311569213867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,2,32,power_law_1.2,0.08391424179077148
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,2,32,power_law_1.2,0.08760191917419433
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,2,32,power_law_1.2,0.0975539207458496
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,2,32,power_law_1.2,0.12359423637390136
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,2,32,power_law_1.2,0.1492684841156006
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,2,32,power_law_1.2,0.07620863914489746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,2,32,power_law_1.2,0.193570556640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,2,32,power_law_1.2,0.28139392852783207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,2,32,power_law_1.2,2.0380364990234376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,32,2,power_law_1.2,1.3005850219726562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,2,32,power_law_1.2,3.153306884765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,2,32,power_law_1.2,0.4394457626342773
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,2,32,power_law_1.2,0.7214412689208984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,4,16,balanced,0.029373440742492672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,2,32,power_law_1.2,0.9598796844482422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,2,32,power_law_1.2,0.3589555358886719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,4,16,balanced,0.02893440008163452
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,4,16,balanced,0.05083775997161866
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,4,16,balanced,0.05933055877685547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,2,32,power_law_1.2,4.194772338867187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,4,16,balanced,0.1282521629333496
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,4,16,balanced,0.13123583793640137
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,4,16,balanced,0.1357222366333008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,4,16,balanced,0.13414527893066405
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,4,16,balanced,0.13786879539489746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,4,16,balanced,0.14012800216674806
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,4,16,balanced,0.14346112251281737
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,4,16,balanced,0.1409164810180664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,2,32,power_law_1.2,1.4480781555175781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,4,16,balanced,0.1450496006011963
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,4,16,balanced,0.14844927787780762
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,4,16,balanced,0.15474687576293947
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,4,16,balanced,0.1694528007507324
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,4,16,balanced,0.17903488159179687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,4,16,balanced,0.2075775909423828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,4,16,balanced,0.07448832035064698
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,4,16,balanced,0.22883071899414062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,4,16,balanced,0.30733823776245117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,4,16,balanced,0.40329471588134763
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,4,16,balanced,0.5333836746215821
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,4,16,balanced,0.029295361042022704
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,4,16,balanced,0.1477849578857422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,4,16,balanced,0.7078899383544922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,4,16,balanced,0.029843199253082275
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,4,16,balanced,0.03744127988815308
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,4,16,balanced,0.05106815814971923
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,4,16,balanced,0.06702208042144775
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,4,16,balanced,0.9790029144287109
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,2,32,power_law_1.2,2.0369165039062502
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,4,16,balanced,0.06990208148956298
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,4,16,balanced,0.07040256023406982
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,4,16,balanced,0.07131519794464111
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,4,16,balanced,0.07274623870849609
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,4,16,balanced,0.07520768165588379
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,4,16,balanced,0.0775923204421997
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,4,16,balanced,0.08141823768615722
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,4,16,balanced,0.07469696044921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,4,16,balanced,0.07777279853820801
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,4,16,balanced,1.2726822662353516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,4,16,balanced,0.08051327705383302
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,4,16,balanced,0.08434944152832032
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,4,16,balanced,0.09575424194335938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,4,16,balanced,0.1081049633026123
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,4,16,balanced,0.1290649604797363
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,4,16,balanced,0.1506713581085205
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,4,16,balanced,0.19961856842041015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,4,16,balanced,0.2392780876159668
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,4,16,balanced,0.029035520553588868
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,4,16,balanced,0.36114944458007814
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,4,16,power_law_1.01,0.07351808071136474
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,4,16,balanced,0.45015422821044926
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,4,16,power_law_1.01,0.0838144016265869
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,4,16,power_law_1.01,0.08506752014160156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,4,16,power_law_1.01,0.08398464202880859
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,4,16,balanced,0.6176153564453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,4,16,power_law_1.01,0.12134911537170409
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,4,16,power_law_1.01,0.12583680152893068
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,4,16,power_law_1.01,0.12896384239196776
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,4,16,power_law_1.01,0.12461952209472656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,4,16,power_law_1.01,0.13195775985717773
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,4,16,balanced,0.8251725006103516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,4,16,power_law_1.01,0.13847552299499513
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,32,2,power_law_1.2,1.2378534698486328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,4,16,power_law_1.01,0.13987199783325194
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,4,16,power_law_1.01,0.14493568420410158
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,4,16,power_law_1.01,0.14538880348205568
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,4,16,power_law_1.01,0.14788607597351072
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,4,16,power_law_1.01,0.1548518371582031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,4,16,power_law_1.01,0.15666560173034666
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,4,16,power_law_1.01,0.2241587257385254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,4,16,power_law_1.01,0.30717567443847654
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,4,16,power_law_1.01,0.3635302352905273
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,4,16,power_law_1.01,0.5154111862182618
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,4,16,power_law_1.01,0.6658380889892579
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,4,16,power_law_1.01,0.049223680496215824
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,4,16,power_law_1.01,0.9129856109619141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,4,16,power_law_1.01,0.05273215770721436
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,4,16,power_law_1.01,0.05326208114624024
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,4,16,power_law_1.01,0.17296127319335936
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,4,16,power_law_1.01,0.1880614471435547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,4,16,power_law_1.01,0.05398784160614014
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,4,16,power_law_1.01,0.06673151969909667
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,4,16,power_law_1.01,0.06766848087310791
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,4,16,power_law_1.01,0.06879104137420654
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,4,16,power_law_1.01,0.07165440082550048
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,4,16,power_law_1.01,0.07438079833984375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,4,16,power_law_1.01,0.07569151878356933
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,4,16,power_law_1.01,0.07731840133666992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,4,16,power_law_1.01,0.07898623943328857
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,4,16,power_law_1.01,0.0779801607131958
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,4,16,power_law_1.01,1.7973107910156252
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,4,16,power_law_1.01,0.0805504035949707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,4,16,power_law_1.01,0.0836019229888916
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,4,16,power_law_1.01,0.09676544189453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,4,16,power_law_1.01,0.12720000267028808
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,4,16,power_law_1.01,0.11218175888061524
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,4,16,power_law_1.01,0.1605619239807129
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,4,16,power_law_1.01,0.06343167781829834
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,4,16,power_law_1.01,1.1879500579833984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,4,16,power_law_1.01,0.18928255081176756
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,4,16,power_law_1.01,2.4769126892089846
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,4,16,power_law_1.01,0.3191641616821289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,4,16,power_law_1.01,0.39370880126953123
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,4,16,power_law_1.2,0.08204159736633301
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,4,16,power_law_1.2,0.0821286392211914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,4,16,power_law_1.01,0.5668915176391602
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,4,16,power_law_1.2,0.0839628791809082
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,4,16,power_law_1.2,0.08394623756408691
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,4,16,power_law_1.01,0.7382988739013672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,4,16,power_law_1.2,0.1149350357055664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,4,16,power_law_1.2,0.12349568367004396
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,4,16,power_law_1.2,0.12611583709716795
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,4,16,power_law_1.2,0.12951040267944336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,4,16,power_law_1.2,0.13134336471557617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,4,16,power_law_1.2,0.13329792022705078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,4,16,power_law_1.2,0.1443276786804199
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,4,16,power_law_1.2,0.14741503715515136
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,4,16,power_law_1.2,0.14808704376220702
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,4,16,power_law_1.2,0.15209856033325195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,4,16,power_law_1.2,0.15876223564147948
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,4,16,power_law_1.2,0.16433023452758788
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,4,16,power_law_1.2,0.17358848571777344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,4,16,power_law_1.2,0.20289024353027343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,4,16,power_law_1.2,0.2572198486328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,4,16,power_law_1.2,0.3189926338195801
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,4,16,power_law_1.2,0.4211891174316406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,4,16,power_law_1.01,1.031910400390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,4,16,power_law_1.2,0.6190195083618164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,4,16,power_law_1.2,0.7716377258300782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,4,16,power_law_1.01,1.3182949829101562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,4,16,power_law_1.2,0.052597761154174805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,4,16,power_law_1.2,0.051774721145629886
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,4,16,power_law_1.2,0.05345920085906982
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,4,16,power_law_1.2,0.05614848136901855
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,4,16,power_law_1.2,0.06614655971527099
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,4,16,power_law_1.2,1.341338806152344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,4,16,power_law_1.2,0.07044479846954346
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,4,16,power_law_1.2,0.06973951816558838
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,4,16,power_law_1.2,0.07035903930664063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,4,16,power_law_1.2,1.5584678649902344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,4,16,power_law_1.2,0.07626495838165284
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,4,16,power_law_1.2,0.08357248306274415
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,4,16,power_law_1.2,0.0765657615661621
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,4,16,power_law_1.2,0.08278400421142577
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,4,16,power_law_1.2,0.08013055801391603
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,4,16,power_law_1.2,0.08583552360534667
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,4,16,power_law_1.2,0.09876992225646972
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,4,16,power_law_1.2,0.12208767890930176
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,4,16,power_law_1.2,0.17776128768920899
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,4,16,power_law_1.2,0.06047999858856201
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,4,16,power_law_1.2,0.2500147247314453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,4,16,power_law_1.2,0.33088897705078124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,4,16,power_law_1.2,0.0728768014907837
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,4,16,power_law_1.2,0.4187968063354492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,8,8,balanced,0.027774720191955565
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,4,16,power_law_1.2,0.6486566162109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,8,8,balanced,0.03762687921524048
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,8,8,balanced,0.05147776126861572
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,4,16,power_law_1.2,0.1362547206878662
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,8,8,balanced,0.05979008197784423
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,4,16,power_law_1.2,0.876262435913086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,8,8,balanced,0.07725696086883545
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,8,8,balanced,0.130699520111084
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,8,8,balanced,0.13387264251708983
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,8,8,balanced,0.13887231826782226
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,8,8,balanced,0.14263168334960938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,8,8,balanced,0.14490495681762697
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,8,8,balanced,0.13670144081115723
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,8,8,balanced,0.14003583908081055
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,8,8,balanced,0.14323967933654785
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,8,8,balanced,0.14828672409057617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,8,8,balanced,0.15282176017761231
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,8,8,balanced,0.15976448059082032
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,8,8,balanced,0.1691200065612793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,8,8,balanced,0.1885478401184082
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,8,8,balanced,0.20760320663452148
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,8,8,balanced,0.2538265609741211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,8,8,balanced,0.29510528564453126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,8,8,balanced,0.40413440704345704
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,8,8,balanced,0.5446156692504883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,8,8,balanced,0.7459878540039062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,8,8,balanced,0.02900223970413208
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,8,8,balanced,0.9807705688476563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,8,8,balanced,0.029224960803985595
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,4,16,power_law_1.2,1.8411712646484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,8,8,balanced,0.031303679943084715
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,8,8,balanced,0.04136320114135742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,8,8,balanced,0.05562880039215088
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,8,8,balanced,1.3837158203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,8,8,balanced,0.07788288116455078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,8,8,balanced,0.08038911819458008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,8,8,balanced,0.08399488449096679
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,8,8,balanced,0.08898816108703614
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,8,8,balanced,0.08164095878601074
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,8,8,balanced,0.08452863693237304
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,8,8,balanced,1.8090623474121095
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,8,8,balanced,0.08690176010131836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,8,8,balanced,0.09133184432983399
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,8,8,balanced,0.09383680343627929
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,8,8,balanced,0.09964287757873536
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,8,8,balanced,0.11036928176879883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,8,8,balanced,0.1304870414733887
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,8,8,balanced,0.1503104019165039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,8,8,balanced,0.18696063995361328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,8,8,balanced,0.22539648056030273
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,8,8,balanced,0.08537343978881837
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,8,8,balanced,0.31876096725463865
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,8,8,balanced,0.3958924865722656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,8,8,balanced,0.586822395324707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,4,16,power_law_1.2,1.352481231689453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,8,8,power_law_1.01,0.0587059211730957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,8,8,power_law_1.01,0.07162752151489257
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,8,8,power_law_1.01,0.07995520114898683
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,8,8,balanced,1.0960205078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,8,8,power_law_1.01,0.07702911853790283
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,8,8,power_law_1.01,0.08028927803039551
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,8,8,power_law_1.01,0.1237286376953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,8,8,power_law_1.01,0.12689663887023925
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,4,16,power_law_1.2,2.2375833129882814
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,8,8,balanced,1.452947235107422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,8,8,power_law_1.01,0.12972543716430665
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,8,8,power_law_1.01,0.13474047660827637
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,8,8,power_law_1.01,0.134716157913208
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,8,8,power_law_1.01,0.14196096420288087
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,8,8,power_law_1.01,0.13810303688049316
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,8,8,balanced,0.7510655975341797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,8,8,power_law_1.01,0.1439795207977295
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,8,8,power_law_1.01,0.13874303817749023
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,8,8,power_law_1.01,0.14923263549804688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,8,8,power_law_1.01,0.15644415855407715
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,8,8,power_law_1.01,0.16879104614257812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,8,8,power_law_1.01,0.18285696029663084
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,8,8,power_law_1.01,0.20112640380859376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,8,8,power_law_1.01,0.26456064224243164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,8,8,power_law_1.01,0.3257958221435547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,8,8,power_law_1.01,0.47845119476318365
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,8,8,power_law_1.01,0.04062592029571534
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,8,8,power_law_1.01,0.5952204895019532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,8,8,power_law_1.01,0.0492300796508789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,8,8,power_law_1.01,0.8443007659912108
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,8,8,power_law_1.01,0.04851712226867676
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,8,8,power_law_1.01,0.05462912082672119
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,8,8,power_law_1.01,0.05436031818389893
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,8,8,power_law_1.01,0.07600512027740478
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,8,8,power_law_1.01,0.07544703960418701
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,8,8,power_law_1.01,0.0789132785797119
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,8,8,power_law_1.01,0.08108927726745604
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,8,8,power_law_1.01,0.08338944435119629
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,8,8,power_law_1.01,0.0870963191986084
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,8,8,power_law_1.01,0.08684160232543944
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,8,8,power_law_1.01,0.0835148811340332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,8,8,power_law_1.01,0.08790911674499512
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,8,8,power_law_1.01,0.08994688034057617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,8,8,power_law_1.01,0.09215231895446777
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,8,8,power_law_1.01,1.5778495788574218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,8,8,power_law_1.01,0.1026700782775879
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,8,8,power_law_1.01,0.1262335968017578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,8,8,power_law_1.01,0.13928319931030272
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,4,16,power_law_1.2,3.080549011230469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,8,8,power_law_1.01,0.18050048828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,8,8,power_law_1.01,0.2219430351257324
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,8,8,power_law_1.01,2.1699749755859377
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,8,8,power_law_1.01,1.0756710052490235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,8,8,power_law_1.01,0.3298688125610352
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,8,8,power_law_1.01,0.3870207977294922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,8,8,power_law_1.01,0.5794303894042969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,8,8,power_law_1.2,0.06400512218475342
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,8,8,power_law_1.2,0.06555391788482666
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,8,8,power_law_1.2,0.06879360198974609
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,8,8,power_law_1.2,0.07631231784820557
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,8,8,power_law_1.01,0.7464179229736329
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,8,8,power_law_1.2,0.12907903671264648
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,8,8,power_law_1.2,0.12447872161865234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,8,8,power_law_1.2,0.13469311714172363
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,8,8,power_law_1.2,0.1410470390319824
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,8,8,power_law_1.01,1.0480973052978517
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,8,8,power_law_1.2,0.13940608024597168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,8,8,power_law_1.2,0.14627327919006347
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,8,8,power_law_1.2,0.15125887870788574
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,8,8,power_law_1.2,0.15806976318359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,8,8,power_law_1.01,1.568133087158203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,8,8,power_law_1.2,0.1846463966369629
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,8,8,power_law_1.2,0.21433216094970703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,8,8,power_law_1.2,0.07443071842193603
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,8,8,power_law_1.2,0.29464576721191404
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,8,8,power_law_1.2,0.3770495986938477
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,8,8,power_law_1.2,0.13861120223999024
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,8,8,power_law_1.2,0.1425100803375244
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,8,8,power_law_1.2,0.5019609451293945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,8,8,power_law_1.2,0.6386547088623047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,8,8,power_law_1.2,0.16582271575927734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,8,8,power_law_1.2,0.04060031890869141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,8,8,power_law_1.2,0.8911373138427734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,8,8,power_law_1.2,0.04564864158630371
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,8,8,power_law_1.2,0.13086463928222655
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,8,8,power_law_1.2,0.04876416206359863
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,8,8,power_law_1.2,0.054941439628601076
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,8,8,power_law_1.2,0.05513855934143066
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,8,8,power_law_1.2,0.07485568046569824
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,8,8,power_law_1.2,1.2895295715332031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,8,8,power_law_1.2,0.07866623878479004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,8,8,power_law_1.2,0.08166272163391113
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,8,8,power_law_1.2,0.08470144271850585
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,8,8,power_law_1.2,0.0869696044921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,8,8,power_law_1.2,0.08354687690734863
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,8,8,power_law_1.2,0.08326399803161622
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,8,8,power_law_1.2,0.08705663681030273
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,8,8,power_law_1.2,0.09037440299987794
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,8,8,power_law_1.2,0.09715456008911133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,8,8,power_law_1.2,0.10687871932983399
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,8,8,power_law_1.2,0.13010432243347167
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,8,8,power_law_1.2,0.15051136016845704
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,8,8,power_law_1.2,0.18955904006958008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,8,8,power_law_1.2,0.24809856414794923
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,8,8,power_law_1.2,0.07536767959594727
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,8,8,power_law_1.2,0.3461875152587891
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,8,8,power_law_1.2,0.4186188888549805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,16,4,balanced,0.02984832048416138
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,16,4,balanced,0.05369088172912597
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,8,8,power_law_1.2,0.6364031982421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,16,4,balanced,0.06492159843444825
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,16,4,balanced,0.08354559898376465
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,8,8,power_law_1.2,0.8705612945556641
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,16,4,balanced,0.09773823738098145
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,16,4,balanced,0.13540096282958985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,16,4,balanced,0.138155517578125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,16,4,balanced,0.14095104217529297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,16,4,balanced,0.1433676815032959
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,16,4,balanced,0.1371020793914795
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,16,4,balanced,0.14014592170715331
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,16,4,balanced,0.1432767963409424
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,16,4,balanced,0.1481817626953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,16,4,balanced,0.15202048301696777
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,16,4,balanced,0.15880576133728028
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,16,4,balanced,0.03659647941589356
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,16,4,balanced,0.16687744140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,16,4,balanced,0.185743350982666
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,16,4,balanced,0.20266111373901366
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,16,4,balanced,0.254911994934082
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,16,4,balanced,0.2896486473083496
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,8,8,power_law_1.2,1.2660467529296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,16,4,balanced,0.4010060882568359
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,16,4,balanced,0.5487846374511719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,16,4,balanced,0.03003904104232788
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,16,4,balanced,0.7530982208251953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,8,8,power_law_1.2,1.7315519714355467
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,16,4,balanced,0.031485440731048586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,16,4,balanced,0.03549312114715576
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,16,4,balanced,0.9907839965820312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,16,4,balanced,0.04764800071716309
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,16,4,balanced,0.07883776187896728
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,16,4,balanced,0.09770624160766601
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,16,4,balanced,0.1011673641204834
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,16,4,balanced,0.1023193645477295
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,16,4,balanced,0.103755521774292
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,16,4,balanced,0.09798015594482422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,16,4,balanced,1.4028787231445312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,16,4,balanced,0.10057727813720703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,16,4,balanced,0.10241408348083496
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,16,4,balanced,0.10925951957702637
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,16,4,balanced,0.12180480003356933
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,16,4,balanced,1.8143052673339846
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,16,4,balanced,0.1425817584991455
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,16,4,balanced,0.15846272468566897
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,16,4,balanced,0.06496384143829345
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,16,4,balanced,0.19576704025268554
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,16,4,balanced,0.23292543411254885
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,16,4,balanced,0.3674291229248047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,16,4,balanced,0.441082878112793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,16,4,balanced,0.10702079772949218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,16,4,balanced,0.6137996673583984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,16,4,balanced,0.11425024032592775
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,16,4,power_law_1.01,0.053288960456848146
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,16,4,balanced,0.8159155273437501
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,16,4,power_law_1.01,0.062105598449707033
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,16,4,power_law_1.01,0.0696230411529541
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,16,4,power_law_1.01,0.07994368076324462
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,16,4,power_law_1.01,0.12863103866577147
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,16,4,balanced,1.1621465301513672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,16,4,power_law_1.01,0.13195008277893067
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,16,4,power_law_1.01,0.13534208297729494
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,16,4,power_law_1.01,0.13818367958068847
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,16,4,power_law_1.01,0.13683199882507324
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,16,4,balanced,1.5395365905761718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,8,8,power_law_1.2,1.8681881713867188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,16,4,power_law_1.01,0.1397926425933838
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,16,4,power_law_1.01,0.06195968151092529
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,16,4,power_law_1.01,0.14663807868957518
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,16,4,power_law_1.01,0.15091327667236326
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,16,4,power_law_1.01,0.1567321586608887
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,16,4,power_law_1.01,0.16877439498901367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,16,4,power_law_1.01,0.19192960739135742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,16,4,power_law_1.01,0.21058687210083007
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,16,4,power_law_1.01,0.25626623153686523
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,16,4,power_law_1.01,0.302663688659668
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,16,4,power_law_1.01,0.14480511665344237
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,16,4,power_law_1.01,0.4457702255249023
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,16,4,power_law_1.01,0.09078144073486329
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,16,4,power_law_1.01,0.5773772811889648
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,16,4,power_law_1.01,0.034666240215301514
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,16,4,power_law_1.01,0.04285696029663086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,16,4,power_law_1.01,0.8424691009521486
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,16,4,power_law_1.01,0.04555520057678223
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,8,8,power_law_1.2,2.621319580078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,16,4,power_law_1.01,0.05395455837249756
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,16,4,power_law_1.01,0.06123775959014892
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,16,4,power_law_1.01,1.0401344299316406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,16,4,power_law_1.01,0.0714892816543579
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,16,4,power_law_1.01,0.09175552368164062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,16,4,power_law_1.01,0.0939084815979004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,16,4,power_law_1.01,0.09787008285522461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,16,4,power_law_1.01,0.10446080207824707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,16,4,power_law_1.01,0.10008319854736328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,16,4,power_law_1.01,0.09722240447998047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,16,4,power_law_1.01,0.09960063934326172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,16,4,power_law_1.01,0.10537216186523438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,16,4,power_law_1.01,1.5504716491699218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,16,4,power_law_1.01,0.10762111663818359
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,16,4,power_law_1.01,0.1124608039855957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,16,4,power_law_1.01,0.12235903739929199
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,16,4,power_law_1.01,0.1470681571960449
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,16,4,power_law_1.01,0.16490495681762696
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,16,4,power_law_1.01,0.21223680496215822
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,16,4,power_law_1.01,2.012079315185547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,16,4,power_law_1.01,0.3584409713745117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,16,4,power_law_1.01,0.45707519531249996
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,16,4,power_law_1.2,0.05701504230499268
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,16,4,power_law_1.2,0.05730303764343262
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,16,4,power_law_1.01,0.6660723114013672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,16,4,power_law_1.2,0.06224512100219727
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,16,4,power_law_1.2,0.07509376049041748
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,16,4,power_law_1.01,0.8906470489501952
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,16,4,power_law_1.2,0.0863974380493164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,16,4,power_law_1.2,0.12925312042236328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,16,4,power_law_1.2,0.13406335830688476
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,16,4,power_law_1.2,0.13635968208312987
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,16,4,power_law_1.2,0.14416383743286132
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,16,4,power_law_1.01,0.26503807067871094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,16,4,power_law_1.2,0.13886591911315918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,16,4,power_law_1.2,0.14057727813720705
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,16,4,power_law_1.2,0.14785663604736327
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,16,4,power_law_1.2,0.15118975639343263
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,16,4,power_law_1.01,1.6008627319335935
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,16,4,power_law_1.2,0.15776000022888184
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,16,4,power_law_1.2,0.16871679306030274
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,16,4,power_law_1.2,0.19329151153564453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,16,4,power_law_1.2,0.21024000167846682
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,16,4,power_law_1.2,0.2653311920166016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,16,4,power_law_1.2,0.3232998275756836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,16,4,power_law_1.2,0.4815552139282227
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,16,4,power_law_1.2,0.1406054401397705
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,16,4,power_law_1.2,0.5942668914794922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,16,4,power_law_1.2,0.0340070390701294
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,16,4,power_law_1.2,0.0687769603729248
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,16,4,power_law_1.2,0.04161920070648194
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,16,4,power_law_1.2,0.8758656311035157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,16,4,power_law_1.2,0.045434880256652835
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,16,4,power_law_1.2,0.05245823860168457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,16,4,power_law_1.2,0.055147519111633306
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,16,4,power_law_1.2,0.09032064437866211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,16,4,power_law_1.2,0.09570303916931153
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,16,4,power_law_1.2,0.09783424377441406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,16,4,power_law_1.2,0.10110719680786133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,16,4,power_law_1.2,0.10180607795715332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,16,4,power_law_1.2,0.09934464454650879
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,16,4,power_law_1.2,0.10055808067321777
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,16,4,power_law_1.2,0.1084224033355713
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,16,4,power_law_1.2,0.11478272438049317
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,16,4,power_law_1.2,0.1253388786315918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,16,4,power_law_1.2,0.14903679847717285
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,16,4,power_law_1.2,0.0745740795135498
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,16,4,power_law_1.2,0.16682495117187499
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,16,4,power_law_1.01,1.1734528350830078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,16,4,power_law_1.2,0.22766080856323243
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,16,4,power_law_1.2,0.2916608047485352
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,16,4,power_law_1.2,0.10600064277648927
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,16,4,power_law_1.2,0.37285633087158204
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,16,4,power_law_1.2,0.46988159179687494
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,16,4,power_law_1.2,1.0652210998535157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,32,2,balanced,0.0291865611076355
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,16,4,power_law_1.2,0.6738521575927734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,32,2,balanced,0.03663232088088989
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,32,2,balanced,0.05520256042480469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,32,2,balanced,0.06939008235931396
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,32,2,balanced,0.0913203239440918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,32,2,balanced,0.1060812759399414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,32,2,balanced,0.10738816261291503
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,32,2,balanced,0.10846464157104492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,32,2,balanced,0.11058688163757324
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,32,2,balanced,0.14161664009094238
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,16,4,power_law_1.2,0.8843417358398437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,32,2,balanced,0.14322431564331056
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,32,2,balanced,0.14659711837768555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,32,2,balanced,0.1535961627960205
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,32,2,balanced,0.15639936447143554
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,32,2,balanced,0.1623948860168457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,32,2,balanced,0.1701273536682129
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,32,2,balanced,0.18982528686523437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,32,2,balanced,0.20767871856689454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,32,2,balanced,0.26595071792602537
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,32,2,balanced,0.30228864669799804
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,32,2,balanced,0.14846207618713378
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,32,2,balanced,0.45441280364990233
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,16,4,power_law_1.2,1.7750912475585938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,32,2,balanced,0.585909767150879
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,32,2,balanced,0.8306380462646483
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,32,2,balanced,0.029630720615386963
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,32,2,balanced,0.03250816106796265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,32,2,balanced,1.0695807647705078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,32,2,balanced,0.039617280960083004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,32,2,balanced,0.053890562057495116
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,32,2,balanced,0.08136063575744629
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,32,2,balanced,0.10196479797363281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,32,2,balanced,0.09752575874328613
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,32,2,balanced,0.09854847908020019
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,32,2,balanced,1.54951171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,32,2,balanced,0.10036735534667969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,32,2,balanced,0.12619775772094727
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,32,2,balanced,0.12781439781188966
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,32,2,balanced,0.13283583641052246
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,32,2,balanced,0.13679103851318358
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,32,2,balanced,0.13787391662597656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,32,2,balanced,0.1433459186553955
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,32,2,balanced,2.037529602050781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,32,2,balanced,0.1497804832458496
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,32,2,balanced,0.16960256576538085
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,32,2,balanced,0.1839039993286133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,16,4,power_law_1.2,1.3166079711914063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,32,2,balanced,0.21753087997436524
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,32,2,balanced,0.24999935150146485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,32,2,balanced,0.40870655059814454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,32,2,balanced,0.4762828826904297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,32,2,balanced,0.13055359840393066
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,32,2,balanced,0.671673583984375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,32,2,power_law_1.01,0.0308787202835083
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,16,4,power_law_1.2,1.6166464233398437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,32,2,power_law_1.01,0.05445248126983643
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,32,2,balanced,0.8971968078613282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,32,2,power_law_1.01,0.07037439823150635
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,32,2,power_law_1.01,0.05934976100921631
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,32,2,power_law_1.01,0.07899136066436767
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,32,2,power_law_1.01,0.09473024368286133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,32,2,power_law_1.01,0.10039551734924317
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,32,2,power_law_1.01,0.09985152244567871
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,32,2,power_law_1.01,0.10476032257080078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,32,2,balanced,1.2562496185302734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,32,2,power_law_1.01,0.13631744384765626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,32,2,power_law_1.01,0.13928447723388673
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,32,2,power_law_1.01,0.1410534381866455
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,32,2,power_law_1.01,0.144716796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,32,2,power_law_1.01,0.15178367614746094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,32,2,power_law_1.01,0.16093439102172852
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,32,2,power_law_1.01,0.17268096923828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,32,2,power_law_1.01,0.19903488159179689
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,32,2,power_law_1.01,0.21981439590454102
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,32,2,power_law_1.01,0.2663795280456543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,32,2,balanced,1.6719859313964842
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,16,4,power_law_1.2,2.2379225158691405
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,32,2,power_law_1.01,0.47675521850585934
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,32,2,power_law_1.01,0.5927347183227539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,32,2,power_law_1.01,0.03227135896682739
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,32,2,power_law_1.01,0.8572351837158203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,32,2,power_law_1.01,0.037785599231719966
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,32,2,power_law_1.01,0.045946879386901854
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,32,2,power_law_1.01,1.1137216186523438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,32,2,power_law_1.01,0.05515776157379151
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,32,2,power_law_1.01,0.08571264266967774
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,32,2,power_law_1.01,0.0874687957763672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,32,2,power_law_1.01,0.09281408309936523
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,32,2,power_law_1.01,1.6158963012695313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,32,2,power_law_1.01,0.11867263793945312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,32,2,power_law_1.01,0.12061568260192872
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,32,2,power_law_1.01,0.1252006435394287
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,32,2,power_law_1.01,0.128602876663208
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,32,2,power_law_1.01,0.15486080169677735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,32,2,power_law_1.01,0.13470848083496095
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,32,2,power_law_1.01,0.3147648048400879
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,32,2,power_law_1.01,0.13510656356811523
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,32,2,power_law_1.01,0.1492633628845215
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,32,2,power_law_1.01,0.1755276870727539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,32,2,power_law_1.01,0.19862783432006836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,32,2,power_law_1.01,0.09078271865844727
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,32,2,power_law_1.01,0.2441472053527832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,32,2,power_law_1.01,0.30444160461425784
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,32,2,power_law_1.01,0.4124889755249024
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,32,2,power_law_1.01,0.5103014373779298
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,32,2,power_law_1.01,0.7209343719482422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,32,2,power_law_1.01,0.13936511993408202
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,32,2,power_law_1.01,2.157560272216797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,32,2,power_law_1.01,0.9301618957519532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,32,2,power_law_1.2,0.031096320152282714
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,32,2,power_law_1.2,0.05295743942260742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,32,2,power_law_1.2,0.05909887790679932
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,32,2,power_law_1.2,0.06851456165313721
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,32,2,power_law_1.2,0.07520383834838867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,32,2,power_law_1.2,0.0940339183807373
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,32,2,power_law_1.2,0.0981004810333252
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,32,2,power_law_1.01,1.331841278076172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,32,2,power_law_1.2,0.1022003173828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,32,2,power_law_1.2,0.10699007987976075
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,32,2,power_law_1.2,0.13714816093444823
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,32,2,power_law_1.2,0.14027903556823731
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,32,2,power_law_1.2,0.14381055831909179
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,32,2,power_law_1.2,0.14611712455749512
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,32,2,power_law_1.2,0.15718655586242675
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,32,2,power_law_1.2,0.16216447830200195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,32,2,power_law_1.01,0.06923647880554198
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,32,2,power_law_1.01,1.7564262390136718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,32,2,power_law_1.2,0.17518848419189453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,32,2,power_law_1.2,0.20116992950439455
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,32,2,power_law_1.2,0.26790912628173824
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,32,2,power_law_1.2,0.3156211280822754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,32,2,power_law_1.2,0.48031360626220704
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,32,2,power_law_1.2,0.6008947372436524
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,32,2,power_law_1.2,0.15353599548339844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,32,2,power_law_1.2,0.03229824066162109
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,32,2,power_law_1.2,0.8723583984375001
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,32,2,power_law_1.2,0.03808511972427368
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,32,2,power_law_1.2,0.04585343837738037
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,32,2,power_law_1.2,0.053223681449890134
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,32,2,power_law_1.2,1.1396812438964843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,32,2,power_law_1.2,0.06297599792480468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,32,2,power_law_1.2,0.08398079872131348
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,32,2,power_law_1.2,0.08755583763122558
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,32,2,power_law_1.2,0.09107071876525878
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,32,2,power_law_1.2,0.09496831893920898
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,32,2,power_law_1.2,0.12103167533874512
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,32,2,power_law_1.2,0.12208512306213379
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,32,2,power_law_1.2,0.12706303596496582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,32,2,power_law_1.2,0.12953984260559082
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,32,2,power_law_1.2,0.13458687782287598
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,32,2,power_law_1.2,0.13604864120483398
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,32,2,power_law_1.2,0.1419276809692383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,32,2,power_law_1.2,0.15292415618896485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,32,2,power_law_1.2,0.22315263748168945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,32,2,power_law_1.2,0.1775334358215332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,32,2,power_law_1.2,0.19879680633544922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,32,2,power_law_1.2,0.24779903411865237
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,32,2,power_law_1.2,0.3054489517211914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,32,2,power_law_1.2,0.4194700622558594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,32,2,power_law_1.2,0.5169881439208984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,1,64,balanced,0.06831103801727295
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,1,64,balanced,0.06808832168579101
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,32,2,power_law_1.2,0.7488396453857422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,1,64,balanced,0.06832255840301513
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,1,64,balanced,0.06805888175964356
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,1,64,balanced,0.07879680156707763
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,1,64,balanced,0.1428326416015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,1,64,balanced,0.20174976348876955
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,1,64,balanced,0.20148607254028322
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,32,2,power_law_1.2,0.948211212158203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,1,64,balanced,0.2030950355529785
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,1,64,balanced,0.20700544357299805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,1,64,balanced,0.20658943176269534
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,1,64,balanced,0.10758527755737304
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,1,64,balanced,0.20925567626953123
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,1,64,balanced,0.21372928619384765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,1,64,balanced,0.21831039428710936
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,1,64,balanced,0.21669376373291013
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,1,64,balanced,0.22744831085205078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,1,64,balanced,0.20118656158447265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,1,64,balanced,0.2413145637512207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,1,64,balanced,0.26154624938964843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,1,64,balanced,0.27777919769287107
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,32,2,power_law_1.2,1.7695654296874999
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,1,64,balanced,0.3704665756225586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,1,64,balanced,0.4287564849853515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,1,64,balanced,0.04062464237213135
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,1,64,balanced,0.5535116958618164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,1,64,balanced,0.040270080566406255
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,1,64,balanced,0.6897740936279296
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,1,64,balanced,0.23094783782958986
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,1,64,balanced,0.040524802207946777
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,1,64,balanced,0.04855807781219482
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,1,64,balanced,0.06488831996917724
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,1,64,balanced,0.9646975708007812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,1,64,balanced,0.08891648292541504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,1,64,balanced,0.09751935958862304
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,1,64,balanced,0.09757568359374999
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,1,64,balanced,0.09855999946594238
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,1,64,balanced,0.09910016059875489
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,1,64,balanced,0.1002393627166748
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,1,64,balanced,0.09960191726684571
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,1,64,balanced,0.102608642578125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,1,64,balanced,0.10456319808959962
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,1,64,balanced,0.10504575729370118
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,1,64,balanced,0.11045503616333008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,1,64,balanced,0.1172492790222168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,1,64,balanced,0.11229056358337401
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,1,64,balanced,0.04068480014801025
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,1,64,balanced,0.1188326358795166
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,1,64,balanced,0.12726271629333497
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,1,64,balanced,0.14420224189758302
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,1,64,balanced,0.15771648406982422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,1,64,balanced,0.18784767150878906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,1,64,balanced,0.25954303741455076
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,32,2,power_law_1.2,1.3557414245605468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,1,64,balanced,0.3115648078918457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,1,64,power_law_1.01,0.14074496269226072
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,1,64,balanced,0.4129190444946289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,1,64,power_law_1.01,0.1402188777923584
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,1,64,power_law_1.01,0.1428236770629883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,1,64,power_law_1.01,0.1415398406982422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,1,64,power_law_1.01,0.1967500877380371
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,1,64,power_law_1.01,0.1844825553894043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,1,64,power_law_1.01,0.18836736679077148
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,32,2,power_law_1.2,1.6329869079589845
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,1,64,power_law_1.01,0.19607551574707033
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,1,64,power_law_1.01,0.20117504119873048
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,1,64,power_law_1.01,0.20269184112548827
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,1,64,power_law_1.01,0.20091520309448244
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,1,64,power_law_1.01,0.20608255386352542
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,1,64,power_law_1.01,0.21147008895874025
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,1,64,power_law_1.01,0.2060799980163574
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,1,64,power_law_1.01,0.21592575073242187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,1,64,power_law_1.01,0.2258687973022461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,1,64,power_law_1.01,0.22282367706298828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,1,64,power_law_1.01,0.2616102409362793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,1,64,power_law_1.01,0.2727807998657227
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,1,64,power_law_1.01,0.361343994140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,1,64,power_law_1.01,0.4441292953491211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,1,64,power_law_1.01,0.6359334564208984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,1,64,power_law_1.01,0.7053670501708984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,1,64,power_law_1.01,0.08804351806640624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,1,64,power_law_1.01,0.08795392036437988
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,1,64,power_law_1.01,0.08908672332763672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,1,64,power_law_1.01,1.103272933959961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,1,64,power_law_1.01,0.08948736190795899
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,1,64,power_law_1.01,0.09537792205810547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,1,64,power_law_1.01,0.08721920013427734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,32,2,power_law_1.2,2.2433433532714844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,1,64,power_law_1.01,0.09023360252380372
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,1,64,power_law_1.01,1.4405375671386718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,1,64,power_law_1.01,0.09802240371704102
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,1,64,power_law_1.01,0.09890303611755372
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,1,64,power_law_1.01,0.10309632301330567
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,1,64,power_law_1.01,0.1013094425201416
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,1,64,power_law_1.01,0.09822848320007324
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,1,64,power_law_1.01,0.10197504043579102
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,1,64,power_law_1.01,0.11148544311523438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,1,64,power_law_1.01,0.11345536231994628
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,1,64,power_law_1.01,0.11583231925964355
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,1,64,power_law_1.01,0.12089856147766113
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,1,64,power_law_1.01,2.1613401794433593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,1,64,power_law_1.01,0.17217023849487306
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,1,64,power_law_1.01,0.20856191635131838
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,1,64,power_law_1.01,0.2677043151855469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,1,64,power_law_1.01,2.3928256225585938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,1,64,power_law_1.01,0.3471142578125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,1,64,power_law_1.01,0.4697983932495117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,1,64,power_law_1.2,0.14198271751403807
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,1,64,power_law_1.01,0.5781875228881835
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,1,64,power_law_1.2,0.14003328323364256
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,1,64,power_law_1.2,0.14332799911499022
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,1,64,power_law_1.2,0.14267264366149904
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,1,64,power_law_1.01,0.11133695602416993
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,1,64,power_law_1.01,0.7887372589111328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,1,64,power_law_1.01,0.1302143955230713
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,1,64,power_law_1.2,0.20037376403808596
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,1,64,power_law_1.2,0.18718080520629882
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,1,64,power_law_1.2,0.19704576492309572
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,1,64,power_law_1.2,0.19783935546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,1,64,power_law_1.2,0.191778564453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,1,64,power_law_1.01,1.1400089263916016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,1,64,power_law_1.2,0.20586624145507812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,1,64,power_law_1.2,0.20801536560058595
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,1,64,power_law_1.2,0.20670719146728517
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,1,64,power_law_1.2,0.21111167907714842
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,1,64,power_law_1.2,0.21968255996704103
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,1,64,power_law_1.2,0.2671795272827148
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,1,64,power_law_1.2,0.23621759414672852
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,1,64,power_law_1.2,0.32300670623779293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,1,64,power_law_1.2,0.3960038375854492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,1,64,power_law_1.2,0.19190399169921873
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,1,64,power_law_1.2,0.5278591918945312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,1,64,power_law_1.2,0.6929843139648437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,1,64,power_law_1.2,0.19422975540161133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,1,64,power_law_1.2,0.08842495918273927
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,1,64,power_law_1.2,0.9241024017333984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,1,64,power_law_1.2,0.08865792274475097
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,1,64,power_law_1.2,0.08918527603149415
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,1,64,power_law_1.2,0.08860416412353515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,1,64,power_law_1.2,0.09603327751159668
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,1,64,power_law_1.2,1.39146240234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,1,64,power_law_1.2,0.09109248161315918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,1,64,power_law_1.2,0.08317055702209472
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,1,64,power_law_1.2,0.0937996768951416
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,1,64,power_law_1.2,0.09567872047424317
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,1,64,power_law_1.2,0.09538559913635254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,1,64,power_law_1.2,0.10166015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,1,64,power_law_1.2,0.10072064399719238
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,1,64,power_law_1.2,0.10256768226623536
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,1,64,power_law_1.2,0.10727423667907714
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,1,64,power_law_1.2,0.11169535636901855
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,1,64,power_law_1.2,0.11515007972717287
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,1,64,power_law_1.2,2.3838783264160157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,1,64,power_law_1.2,0.12517888069152833
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,1,64,power_law_1.2,0.14317440032958983
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,1,64,power_law_1.2,0.2038118362426758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,1,64,power_law_1.2,0.22997760772705078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,1,64,power_law_1.2,0.32088321685791016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,1,64,power_law_1.2,0.39326847076416016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,1,64,power_law_1.2,0.11288448333740235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,2,32,balanced,0.05664512157440186
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,1,64,power_law_1.2,3.510941467285156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,2,32,balanced,0.05631743907928467
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,1,64,power_law_1.2,3.6458432006835935
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,2,32,balanced,0.05925119876861572
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,2,32,balanced,0.0637337589263916
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,2,32,balanced,0.07563519954681397
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,1,64,power_law_1.2,0.9811199951171876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,2,32,balanced,0.1146406364440918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,2,32,balanced,0.1436569595336914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,2,32,balanced,0.21336448669433591
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,2,32,balanced,0.2150899124145508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,1,64,power_law_1.2,1.2104358673095703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,2,32,balanced,0.2155558395385742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,2,32,balanced,0.22197759628295896
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,2,32,balanced,0.22503551483154297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,1,64,power_law_1.2,1.5027366638183595
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,2,32,balanced,0.2258483123779297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,2,32,balanced,0.22575103759765627
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,1,64,power_law_1.2,0.7415795135498047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,2,32,balanced,0.23227264404296877
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,2,32,balanced,0.2628006362915039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,2,32,balanced,0.24872320175170898
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,2,32,balanced,0.2781913566589355
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,2,32,balanced,0.3128268814086914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,2,32,balanced,0.39970687866210936
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,2,32,balanced,0.21737087249755857
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,2,32,balanced,0.03273983955383301
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,2,32,balanced,0.48302848815917965
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,2,32,balanced,0.032463359832763675
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,2,32,balanced,0.22170495986938477
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,2,32,balanced,0.6339904022216797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,2,32,balanced,0.038656001091003415
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,2,32,balanced,0.22717056274414063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,2,32,balanced,0.04914688110351563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,2,32,balanced,0.06703487873077393
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,2,32,balanced,0.8087155151367188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,2,32,balanced,0.08917375564575195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,2,32,balanced,0.09938688278198242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,2,32,balanced,0.09847295761108399
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,2,32,balanced,0.10030336380004883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,2,32,balanced,0.10295424461364747
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,2,32,balanced,0.10710528373718262
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,2,32,balanced,0.10929408073425292
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,2,32,balanced,0.11258367538452148
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,2,32,balanced,0.1071667194366455
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,2,32,balanced,0.1134284782409668
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,2,32,balanced,0.11928319931030273
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,2,32,balanced,0.13347583770751953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,2,32,balanced,0.15073023796081544
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,2,32,balanced,0.17514495849609374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,2,32,balanced,0.2006771278381348
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,2,32,balanced,0.03270528078079223
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,2,32,balanced,0.24952959060668944
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,2,32,balanced,0.3393471908569336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,2,32,power_law_1.01,0.11510272026062011
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,2,32,balanced,0.10350208282470703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,2,32,power_law_1.01,0.14047231674194335
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,2,32,balanced,0.4357299041748047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,2,32,balanced,0.10262911796569825
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,2,32,balanced,1.1232025909423828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,2,32,balanced,0.5813529586791992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,2,32,power_law_1.01,0.14220800399780273
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,2,32,power_law_1.01,0.14058112144470214
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,2,32,power_law_1.01,0.21099264144897462
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,2,32,power_law_1.01,0.2012019157409668
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,2,32,power_law_1.01,0.18442239761352539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,2,32,power_law_1.01,0.19854080200195312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,2,32,power_law_1.01,0.20431488037109374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,2,32,power_law_1.01,0.21110015869140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,2,32,power_law_1.01,0.2138649559020996
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,2,32,power_law_1.01,0.2212326431274414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,2,32,power_law_1.01,0.22075775146484372
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,2,32,power_law_1.01,0.2267302322387695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,2,32,power_law_1.01,0.22596607208251954
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,2,32,power_law_1.01,0.2340377616882324
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,2,32,power_law_1.01,0.2440268707275391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,2,32,power_law_1.01,0.26603776931762696
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,2,32,power_law_1.01,0.3304102325439453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,2,32,power_law_1.01,0.3754534530639648
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,2,32,power_law_1.01,0.5102937698364258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,2,32,power_law_1.01,0.06502912044525147
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,2,32,power_law_1.01,0.6633971405029297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,2,32,power_law_1.01,0.20987007141113284
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,2,32,power_law_1.01,0.08672767639160156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,2,32,power_law_1.01,0.849587173461914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,2,32,power_law_1.01,0.08763903617858887
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,2,32,power_law_1.01,0.08781184196472168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,2,32,power_law_1.01,0.09676544189453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,2,32,power_law_1.01,1.140222702026367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,2,32,power_law_1.01,0.08982144355773926
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,2,32,power_law_1.01,0.09330944061279298
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,2,32,power_law_1.01,0.09703935623168945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,2,32,power_law_1.01,0.09845248222351075
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,2,32,power_law_1.01,1.586571502685547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,2,32,power_law_1.01,0.10370176315307618
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,2,32,power_law_1.01,0.10646528244018554
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,2,32,power_law_1.01,0.11224448204040527
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,2,32,power_law_1.01,0.1127347183227539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,2,32,power_law_1.01,0.10643199920654296
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,2,32,power_law_1.01,0.11213055610656739
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,2,32,power_law_1.01,0.12209407806396484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,2,32,power_law_1.01,0.13310336112976073
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,2,32,power_law_1.01,0.09290752410888672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,2,32,power_law_1.01,0.20363391876220702
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,2,32,power_law_1.01,0.2687104034423828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,2,32,power_law_1.01,0.10147711753845215
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,2,32,power_law_1.01,0.36666881561279296
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,2,32,power_law_1.01,0.4821632003784179
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,2,32,power_law_1.2,0.1152729606628418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,2,32,power_law_1.01,0.6002201461791993
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,2,32,power_law_1.01,2.2160908508300783
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,2,32,power_law_1.2,0.1408179187774658
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,2,32,power_law_1.01,0.16437376022338868
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,2,32,power_law_1.01,0.8202124786376952
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,2,32,power_law_1.2,0.1425279998779297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,2,32,power_law_1.2,0.14386303901672365
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,2,32,power_law_1.2,0.19144832611083984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,2,32,power_law_1.2,0.1988492774963379
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,2,32,power_law_1.2,0.2025267219543457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,2,32,power_law_1.2,0.21389823913574219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,2,32,power_law_1.2,0.21799039840698242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,2,32,power_law_1.2,0.20987903594970705
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,2,32,power_law_1.2,0.20871679306030275
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,2,32,power_law_1.2,0.21961343765258787
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,2,32,power_law_1.2,0.22280576705932614
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,2,32,power_law_1.2,0.22622463226318362
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,2,32,power_law_1.2,0.141146879196167
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,2,32,power_law_1.2,0.23637760162353513
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,2,32,power_law_1.2,0.2580851173400879
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,2,32,power_law_1.2,0.27846527099609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,2,32,power_law_1.2,0.1993152046203613
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,2,32,power_law_1.2,0.3693734359741211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,2,32,power_law_1.01,1.1460300445556642
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,2,32,power_law_1.2,0.5630720138549805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,2,32,power_law_1.2,0.0704089593887329
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,2,32,power_law_1.2,0.8376268768310547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,2,32,power_law_1.2,0.08650624275207519
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,2,32,power_law_1.2,0.08763008117675782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,2,32,power_law_1.2,0.08796159744262695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,2,32,power_law_1.2,1.222586898803711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,2,32,power_law_1.2,0.09555456161499024
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,2,32,power_law_1.2,0.09523455619812012
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,2,32,power_law_1.2,1.3922866821289062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,2,32,power_law_1.2,0.42431999206542964
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,2,32,power_law_1.2,0.09020159721374513
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,2,32,power_law_1.2,0.0949888038635254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,2,32,power_law_1.2,0.09717760086059571
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,2,32,power_law_1.2,0.10016384124755859
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,2,32,power_law_1.2,0.10171392440795898
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,2,32,power_law_1.2,0.10762240409851073
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,2,32,power_law_1.2,0.11063936233520508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,2,32,power_law_1.2,0.10228992462158204
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,2,32,power_law_1.2,0.10713600158691405
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,2,32,power_law_1.2,0.11418751716613769
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,2,32,power_law_1.2,0.12728063583374022
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,2,32,power_law_1.2,0.14722432136535643
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,2,32,power_law_1.2,0.21169919967651368
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,2,32,power_law_1.2,0.24252799987792967
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,2,32,power_law_1.2,0.3073510360717774
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,2,32,power_law_1.2,0.32682880401611325
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,2,32,power_law_1.2,0.09609984397888184
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,4,16,balanced,0.03170047998428345
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,4,16,balanced,0.03177472114562988
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,2,32,power_law_1.2,2.976093444824219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,2,32,power_law_1.2,0.5651865768432617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,4,16,balanced,0.05540095806121827
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,4,16,balanced,0.06294911861419678
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,4,16,balanced,0.07902592182159424
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,4,16,balanced,0.11129599571228029
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,4,16,balanced,0.14827391624450684
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,2,32,power_law_1.2,0.832968978881836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,4,16,balanced,0.20838783264160154
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,4,16,balanced,0.21015039443969727
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,2,32,power_law_1.2,2.358458862304688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,4,16,balanced,0.21012863159179687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,4,16,balanced,0.21463552474975586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,4,16,balanced,0.21644031524658205
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,4,16,balanced,0.22012159347534183
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,4,16,balanced,0.21475456237792967
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,2,32,power_law_1.2,1.1800614166259766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,4,16,balanced,0.2190771293640137
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,4,16,balanced,0.22589183807373048
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,4,16,balanced,0.23860607147216797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,4,16,balanced,0.2501580810546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,4,16,balanced,0.27849727630615234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,2,32,power_law_1.2,1.4775155639648436
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,4,16,balanced,0.3072870445251465
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,4,16,balanced,0.46722942352294916
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,4,16,balanced,0.6003289413452149
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,4,16,balanced,0.0315391993522644
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,4,16,balanced,0.031703040599823
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,4,16,balanced,0.0339468789100647
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,4,16,balanced,0.21933311462402344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,4,16,balanced,0.04046463966369629
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,4,16,balanced,0.05388160228729248
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,4,16,balanced,1.0417919921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,4,16,balanced,0.07315711975097657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,4,16,balanced,0.0954150390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,4,16,balanced,0.10935423851013185
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,4,16,balanced,0.34932479858398435
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,4,16,balanced,0.1104524803161621
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,4,16,balanced,0.11083776473999023
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,4,16,balanced,0.11597951889038086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,4,16,balanced,0.11960063934326173
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,4,16,balanced,0.11263360023498534
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,4,16,balanced,0.1161945629119873
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,4,16,balanced,0.11824895858764647
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,4,16,balanced,0.12422528266906738
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,4,16,balanced,0.13645055770874023
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,4,16,balanced,0.1505459213256836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,4,16,balanced,0.7954534149169922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,4,16,balanced,0.17083648681640623
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,4,16,balanced,0.1956787109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,4,16,balanced,0.23949440002441405
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,4,16,balanced,0.2846899223327637
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,4,16,balanced,0.37025535583496094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,4,16,power_law_1.01,0.07928448200225831
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,4,16,power_law_1.01,0.10814720153808595
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,4,16,balanced,0.11683199882507325
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,4,16,balanced,1.415548095703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,4,16,power_law_1.01,0.13330944061279298
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,4,16,balanced,0.6838015747070313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,4,16,power_law_1.01,0.140447998046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,4,16,power_law_1.01,0.14043647766113282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,4,16,power_law_1.01,0.2077030372619629
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,4,16,power_law_1.01,0.184768009185791
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,4,16,balanced,0.9135743713378905
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,4,16,power_law_1.01,0.19167232513427734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,4,16,power_law_1.01,0.2015193557739258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,4,16,power_law_1.01,0.19397375106811524
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,4,16,power_law_1.01,0.19910655975341798
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,4,16,power_law_1.01,0.20605184555053713
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,4,16,power_law_1.01,0.20103935241699217
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,4,16,power_law_1.01,0.21539327621459964
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,4,16,power_law_1.01,0.21856128692626955
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,4,16,power_law_1.01,0.223767032623291
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,4,16,power_law_1.01,0.23110015869140624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,4,16,power_law_1.01,0.24995071411132813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,4,16,power_law_1.01,0.30646528244018556
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,4,16,power_law_1.01,0.3598643112182617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,4,16,power_law_1.01,0.49414016723632814
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,4,16,power_law_1.01,0.198657283782959
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,4,16,power_law_1.01,0.6278873443603515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,4,16,power_law_1.01,0.05092607975006104
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,4,16,power_law_1.01,0.8156748962402343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,4,16,power_law_1.01,0.07021696090698241
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,4,16,power_law_1.01,0.09086591720581054
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,4,16,balanced,0.5110489654541015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,4,16,power_law_1.01,0.0909119987487793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,4,16,power_law_1.01,1.0364608001708984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,4,16,power_law_1.01,0.09429887771606446
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,4,16,power_law_1.01,0.10885760307312012
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,4,16,power_law_1.01,0.09819775581359863
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,4,16,power_law_1.01,0.10501503944396973
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,4,16,power_law_1.01,1.3919705200195314
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,4,16,power_law_1.01,0.10996607780456542
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,4,16,power_law_1.01,0.11666175842285156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,4,16,power_law_1.01,0.11620223999023438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,4,16,power_law_1.01,0.10621824264526367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,4,16,power_law_1.01,0.11424639701843262
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,4,16,power_law_1.01,0.11472000122070311
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,4,16,power_law_1.01,0.12230400085449218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,4,16,power_law_1.01,0.14102272033691404
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,4,16,power_law_1.01,1.9530035400390624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,4,16,power_law_1.01,0.15767040252685546
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,4,16,power_law_1.01,0.18241151809692382
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,4,16,power_law_1.01,0.2199193572998047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,4,16,power_law_1.01,0.296313591003418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,4,16,power_law_1.01,0.10547455787658691
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,4,16,power_law_1.01,0.3564070510864258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,4,16,power_law_1.2,0.07915135860443115
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,4,16,power_law_1.01,0.4805120086669922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,4,16,power_law_1.2,0.10812416076660156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,4,16,power_law_1.01,0.6515084838867187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,4,16,power_law_1.2,0.13302144050598144
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,4,16,power_law_1.2,0.1450496006011963
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,4,16,power_law_1.2,0.20161407470703124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,4,16,power_law_1.01,0.9818598175048828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,4,16,power_law_1.2,0.19522176742553712
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,4,16,power_law_1.2,0.2005414390563965
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,4,16,power_law_1.01,1.1624947357177735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,4,16,power_law_1.2,0.19449600219726562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,4,16,power_law_1.01,0.10445952415466309
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,4,16,power_law_1.2,0.20453887939453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,4,16,power_law_1.2,0.19990655899047852
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,4,16,power_law_1.2,0.20568832397460937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,4,16,power_law_1.2,0.20771839141845705
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,4,16,power_law_1.2,0.21306112289428708
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,4,16,power_law_1.2,0.21590015411376956
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,4,16,power_law_1.2,0.22167936325073243
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,4,16,power_law_1.2,0.2357004737854004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,4,16,power_law_1.2,0.2637977600097656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,4,16,power_law_1.2,0.3263014221191406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,4,16,power_law_1.2,0.3996236801147461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,4,16,power_law_1.2,0.5634112167358398
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,4,16,power_law_1.2,0.20890111923217775
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,4,16,power_law_1.2,0.1363929557800293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,4,16,power_law_1.2,0.050943999290466314
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,4,16,power_law_1.2,0.9516697692871094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,4,16,power_law_1.2,0.07015168190002441
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,4,16,power_law_1.2,0.08934016227722168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,4,16,power_law_1.2,0.08682623863220215
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,4,16,power_law_1.2,0.09390336036682129
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,4,16,power_law_1.2,0.10280320167541504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,4,16,power_law_1.2,0.6930508422851562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,4,16,power_law_1.2,0.10210816383361816
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,4,16,power_law_1.2,0.10617728233337402
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,4,16,power_law_1.2,0.10470272064208985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,4,16,power_law_1.2,1.834141387939453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,4,16,power_law_1.2,0.10716416358947753
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,4,16,power_law_1.2,0.10843135833740233
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,4,16,power_law_1.2,0.11751808166503905
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,4,16,power_law_1.2,0.1104524803161621
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,4,16,power_law_1.2,0.11551487922668457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,4,16,power_law_1.2,0.11792896270751954
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,4,16,power_law_1.2,0.11416064262390138
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,4,16,power_law_1.2,0.12302592277526855
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,4,16,power_law_1.2,0.1483123207092285
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,4,16,power_law_1.2,1.2660211181640624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,4,16,power_law_1.2,0.2056268882751465
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,4,16,power_law_1.2,2.3948045349121094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,4,16,power_law_1.2,0.2741836738586426
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,4,16,power_law_1.2,0.32265472412109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,4,16,power_law_1.2,0.43195137023925784
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,8,8,balanced,0.03382911920547486
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,4,16,power_law_1.2,0.4874163055419922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,8,8,balanced,0.05797247886657715
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,8,8,balanced,0.08288127899169921
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,4,16,power_law_1.2,0.7241522979736328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,8,8,balanced,0.12180095672607423
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,8,8,balanced,0.15475071907043456
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,4,16,power_law_1.2,0.16744831085205078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,8,8,balanced,0.20901887893676757
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,8,8,balanced,0.20709375381469725
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,8,8,balanced,0.21220991134643558
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,8,8,balanced,0.20748672485351563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,8,8,balanced,0.21036544799804688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,8,8,balanced,0.2133772850036621
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,8,8,balanced,0.03286272048950195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,4,16,power_law_1.2,1.4003135681152343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,8,8,balanced,0.22093183517456055
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,8,8,balanced,0.22743551254272462
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,8,8,balanced,0.23105024337768554
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,8,8,balanced,0.239998722076416
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,8,8,balanced,0.0688640022277832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,8,8,balanced,0.26136959075927735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,8,8,balanced,0.28116607666015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,8,8,balanced,0.37094783782958984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,8,8,balanced,0.4582847976684571
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,4,16,power_law_1.2,1.029032974243164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,8,8,balanced,0.6279411315917969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,8,8,balanced,0.03242111921310425
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,8,8,balanced,0.033172481060028076
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,8,8,balanced,0.036634879112243654
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,8,8,balanced,0.04779903888702393
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,8,8,balanced,0.8698931121826172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,8,8,balanced,0.05994368076324463
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,8,8,balanced,0.08381823539733887
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,8,8,balanced,0.11101823806762696
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,8,8,balanced,0.13247488021850587
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,8,8,balanced,1.112590103149414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,8,8,balanced,0.13415167808532716
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,8,8,balanced,0.13668479919433593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,8,8,balanced,0.13369855880737305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,8,8,balanced,0.13652607917785645
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,8,8,balanced,0.14096256256103518
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,8,8,balanced,0.14782976150512694
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,8,8,balanced,0.15275775909423828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,8,8,balanced,1.5336601257324218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,8,8,balanced,0.16515071868896486
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,8,8,balanced,0.18386688232421874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,8,8,balanced,0.20582271575927735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,8,8,balanced,0.247139835357666
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,8,8,balanced,0.28781824111938475
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,8,8,balanced,0.37206657409667965
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,8,8,balanced,2.040518341064453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,8,8,balanced,0.4571980667114258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,8,8,balanced,0.327217903137207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,8,8,balanced,0.6188544082641602
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,8,8,power_law_1.01,0.07170432090759278
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,8,8,power_law_1.01,0.07795839786529542
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,8,8,balanced,0.8515455627441406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,8,8,power_law_1.01,0.1047372817993164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,8,8,power_law_1.01,0.12221823692321779
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,8,8,power_law_1.01,0.13269760131835937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,8,8,power_law_1.01,0.13583616256713865
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,8,8,balanced,1.1799052429199217
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,8,8,power_law_1.01,0.20020095825195314
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,8,8,power_law_1.01,0.1835737609863281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,8,8,balanced,0.1306060791015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,8,8,power_law_1.01,0.18731136322021485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,8,8,power_law_1.01,0.20272127151489258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,8,8,balanced,1.583276824951172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,8,8,power_law_1.01,0.20410240173339841
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,8,8,power_law_1.01,0.20720767974853516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,8,8,power_law_1.01,0.10836607933044433
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,8,8,power_law_1.01,0.20745983123779296
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,8,8,power_law_1.01,0.21851264953613284
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,8,8,power_law_1.01,0.23189888000488282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,8,8,power_law_1.01,0.25863296508789063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,8,8,power_law_1.01,0.30819583892822267
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,8,8,power_law_1.01,0.3639923095703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,8,8,power_law_1.01,0.4929958343505859
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,8,8,power_law_1.01,0.19171072006225587
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,8,8,power_law_1.01,0.18777856826782227
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,8,8,power_law_1.01,0.6077145767211914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,8,8,power_law_1.01,0.04303743839263916
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,8,8,power_law_1.01,0.05602047920227051
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,8,8,power_law_1.01,0.8245030212402344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,8,8,power_law_1.01,0.07847167968749999
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,8,8,power_law_1.01,0.07528319835662842
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,8,8,power_law_1.01,0.0884108829498291
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,8,8,power_law_1.01,1.075973129272461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,8,8,power_law_1.01,0.09746560096740722
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,8,8,power_law_1.01,0.10418047904968261
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,8,8,power_law_1.01,0.13144960403442382
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,8,8,power_law_1.01,0.12324095726013184
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,8,8,power_law_1.01,0.1190924835205078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,8,8,power_law_1.01,1.5382745361328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,8,8,power_law_1.01,0.12906880378723146
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,8,8,power_law_1.01,0.12911104202270507
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,8,8,power_law_1.01,0.12438528060913086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,8,8,power_law_1.01,0.12916223526000975
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,8,8,power_law_1.01,0.13799424171447755
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,8,8,power_law_1.01,0.1654719924926758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,8,8,power_law_1.01,0.18325759887695312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,8,8,power_law_1.01,0.22732799530029296
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,8,8,power_law_1.01,0.2653683280944824
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,8,8,power_law_1.01,0.3582758331298828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,8,8,power_law_1.01,2.183997497558594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,8,8,power_law_1.01,0.45537025451660157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,8,8,power_law_1.2,0.07133440017700196
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,8,8,power_law_1.2,0.07789311885833741
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,8,8,power_law_1.01,0.13378944396972656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,8,8,power_law_1.01,0.7623628997802735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,8,8,power_law_1.2,0.09403519630432129
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,8,8,power_law_1.01,0.14201600074768067
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,8,8,power_law_1.2,0.09754112243652344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,8,8,power_law_1.2,0.11453439712524413
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,8,8,power_law_1.2,0.14159487724304198
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,8,8,power_law_1.2,0.14755200386047362
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,8,8,power_law_1.01,1.1651302337646485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,8,8,power_law_1.2,0.20622207641601564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,8,8,power_law_1.2,0.18891904830932618
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,8,8,power_law_1.2,0.19231231689453127
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,8,8,power_law_1.2,0.1853990364074707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,8,8,power_law_1.2,0.1868544006347656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,8,8,power_law_1.01,0.6180659103393554
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,8,8,power_law_1.2,0.20679807662963867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,8,8,power_law_1.2,0.20632064819335935
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,8,8,power_law_1.2,0.20792959213256834
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,8,8,power_law_1.2,0.20769279479980468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,8,8,power_law_1.2,0.21739519119262693
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,8,8,power_law_1.2,0.23780096054077146
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,8,8,power_law_1.2,0.263507194519043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,8,8,power_law_1.2,0.3015692710876465
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,8,8,power_law_1.2,0.38250495910644533
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,8,8,power_law_1.2,0.5296409606933594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,8,8,power_law_1.2,0.042593278884887696
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,8,8,power_law_1.2,0.6421401977539063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,8,8,power_law_1.2,0.05585663795471192
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,8,8,power_law_1.01,1.4695706176757812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,8,8,power_law_1.2,0.06609663963317872
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,8,8,power_law_1.2,0.9492095947265625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,8,8,power_law_1.2,0.06993919849395752
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,8,8,power_law_1.2,0.0832307243347168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,8,8,power_law_1.2,1.1802291107177734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,8,8,power_law_1.2,0.10373120307922364
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,8,8,power_law_1.2,0.10814335823059082
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,8,8,power_law_1.2,0.12555904388427735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,8,8,power_law_1.2,0.12790783882141113
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,8,8,power_law_1.2,0.11652480125427248
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,8,8,power_law_1.2,0.13435775756835938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,8,8,power_law_1.2,0.1333401584625244
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,8,8,power_law_1.2,1.7443801879882812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,8,8,power_law_1.2,0.13477120399475098
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,8,8,power_law_1.2,0.13099136352539062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,8,8,power_law_1.2,0.14156160354614258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,8,8,power_law_1.2,0.16818559646606446
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,8,8,power_law_1.2,0.19054079055786133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,8,8,power_law_1.2,0.23378303527832028
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,8,8,power_law_1.2,0.2829017639160156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,8,8,power_law_1.2,0.37327232360839846
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,8,8,power_law_1.2,0.12659584045410158
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,8,8,power_law_1.2,2.6211083984375003
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,8,8,power_law_1.2,0.47178497314453127
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,8,8,power_law_1.2,0.13111167907714844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,16,4,balanced,0.03072256088256836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,16,4,balanced,0.03245696067810059
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,16,4,balanced,0.05759744167327881
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,16,4,balanced,0.06965504169464112
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,16,4,balanced,0.08838272094726562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,8,8,power_law_1.2,0.7930854034423829
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,16,4,balanced,0.12003071784973145
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,16,4,balanced,0.1556544017791748
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,16,4,balanced,0.15954303741455078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,16,4,balanced,0.16013952255249025
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,16,4,balanced,0.16161279678344725
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,16,4,balanced,0.21023359298706054
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,16,4,balanced,0.2138688087463379
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,16,4,balanced,0.2154969596862793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,16,4,balanced,0.22097919464111326
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,8,8,power_law_1.2,1.198551025390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,16,4,balanced,0.22716159820556642
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,16,4,balanced,0.23225984573364258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,16,4,balanced,0.24011648178100584
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,16,4,balanced,0.2583500862121582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,8,8,power_law_1.2,1.5461260986328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,8,8,power_law_1.2,0.6534783935546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,16,4,balanced,0.3175296020507813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,16,4,balanced,0.3610790252685547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,16,4,balanced,0.44717567443847656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,16,4,balanced,0.6265407943725586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,16,4,balanced,0.030590720176696777
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,16,4,balanced,0.0321727991104126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,16,4,balanced,0.037004799842834474
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,16,4,balanced,0.8514969635009766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,16,4,balanced,0.04992256164550781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,16,4,balanced,0.06906879901885986
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,16,4,balanced,1.140838394165039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,16,4,balanced,0.09902463912963867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,16,4,balanced,0.27717119216918945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,16,4,balanced,0.13452287673950197
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,16,4,balanced,0.13622528076171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,16,4,balanced,0.13812992095947266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,16,4,balanced,0.17356672286987304
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,16,4,balanced,0.17588224411010742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,16,4,balanced,0.17953023910522461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,16,4,balanced,0.18385280609130858
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,16,4,balanced,1.5773388671875002
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,16,4,balanced,0.18883071899414064
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,16,4,balanced,0.20036224365234373
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,16,4,balanced,0.22091136932373048
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,16,4,balanced,0.24179199218749997
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,16,4,balanced,0.28297344207763675
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,16,4,balanced,0.31905792236328123
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,16,4,balanced,2.104897308349609
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,16,4,balanced,0.39874431610107425
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,16,4,balanced,0.4766169738769531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,16,4,balanced,0.13684351921081545
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,16,4,power_law_1.01,0.05475327968597412
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,16,4,balanced,0.6340121459960938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,16,4,power_law_1.01,0.0644927978515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,16,4,power_law_1.01,0.07905920028686524
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,16,4,power_law_1.01,0.08181119918823242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,16,4,balanced,0.8972889709472657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,16,4,power_law_1.01,0.10279040336608887
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,16,4,power_law_1.01,0.11721343994140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,16,4,power_law_1.01,0.12040831565856933
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,16,4,power_law_1.01,0.13806079864501952
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,16,4,balanced,1.214677734375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,16,4,power_law_1.01,0.14381567955017088
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,16,4,power_law_1.01,0.18922367095947265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,16,4,power_law_1.01,0.20118783950805663
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,16,4,power_law_1.01,0.20374656677246095
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,16,4,power_law_1.01,0.20669439315795896
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,16,4,power_law_1.01,0.20811647415161133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,16,4,balanced,1.6459494018554686
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,16,4,power_law_1.01,0.222926082611084
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,16,4,balanced,0.1916262435913086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,16,4,power_law_1.01,0.23902463912963867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,16,4,power_law_1.01,0.2672640037536621
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,16,4,power_law_1.01,0.3180953598022461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,16,4,power_law_1.01,0.36580734252929686
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,16,4,power_law_1.01,0.13926015853881837
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,16,4,power_law_1.01,0.5055244827270509
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,16,4,power_law_1.01,0.6283135986328124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,16,4,power_law_1.01,0.03503232002258301
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,16,4,power_law_1.01,0.046682882308959964
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,16,4,power_law_1.01,0.21200639724731446
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,16,4,power_law_1.01,0.900939483642578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,16,4,power_law_1.01,0.06278016090393065
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,16,4,power_law_1.01,0.06714879989624023
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,16,4,power_law_1.01,0.08224639892578126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,16,4,power_law_1.01,0.09581952095031739
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,16,4,power_law_1.01,0.09982720375061035
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,16,4,power_law_1.01,0.11804927825927733
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,16,4,power_law_1.01,0.12477567672729492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,16,4,power_law_1.01,0.15242752075195312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,16,4,power_law_1.01,1.5549349975585938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,16,4,power_law_1.01,0.1690662384033203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,16,4,power_law_1.01,0.16820352554321288
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,16,4,power_law_1.01,0.16315391540527344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,16,4,power_law_1.01,0.17198976516723632
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,16,4,power_law_1.01,0.18408575057983398
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,16,4,power_law_1.01,2.0656332397460937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,16,4,power_law_1.01,0.20906496047973633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,16,4,power_law_1.01,0.22990335464477538
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,16,4,power_law_1.01,0.27526527404785156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,16,4,power_law_1.01,0.16168960571289062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,16,4,power_law_1.01,0.31658111572265624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,16,4,power_law_1.01,0.40700801849365237
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,16,4,power_law_1.01,0.5226265716552734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,16,4,power_law_1.01,0.7121958160400391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,16,4,power_law_1.2,0.0549235200881958
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,16,4,power_law_1.01,1.1524006652832033
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,16,4,power_law_1.01,0.12185088157653809
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,16,4,power_law_1.2,0.06540671825408936
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,16,4,power_law_1.2,0.07371391773223876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,16,4,power_law_1.2,0.08302592277526856
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,16,4,power_law_1.2,0.0981875228881836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,16,4,power_law_1.2,0.1160217571258545
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,16,4,power_law_1.01,1.261190414428711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,16,4,power_law_1.2,0.12750335693359377
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,16,4,power_law_1.2,0.1417625617980957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,16,4,power_law_1.2,0.14553215980529785
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,16,4,power_law_1.2,0.1486847972869873
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,16,4,power_law_1.2,0.19352447509765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,16,4,power_law_1.2,0.19548927307128908
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,16,4,power_law_1.2,0.20196992874145508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,16,4,power_law_1.2,0.21229183197021487
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,16,4,power_law_1.2,0.2016921615600586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,16,4,power_law_1.01,0.8720358276367188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,16,4,power_law_1.2,0.2095564842224121
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,16,4,power_law_1.2,0.21967487335205077
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,16,4,power_law_1.2,0.24351360321044924
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,16,4,power_law_1.2,0.2741350364685059
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,16,4,power_law_1.2,0.3163520050048828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,16,4,power_law_1.2,0.38844287872314454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,16,4,power_law_1.2,0.5302067184448243
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,16,4,power_law_1.2,0.03505664110183716
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,16,4,power_law_1.2,0.6670118713378906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,16,4,power_law_1.01,1.6373452758789064
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,16,4,power_law_1.2,0.04663167953491211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,16,4,power_law_1.2,0.9512486267089845
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,16,4,power_law_1.2,0.05658624172210693
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,16,4,power_law_1.2,0.06545663833618164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,16,4,power_law_1.2,0.08255871772766113
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,16,4,power_law_1.2,1.1948365020751954
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,16,4,power_law_1.2,0.10667776107788085
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,16,4,power_law_1.2,0.12545280456542968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,16,4,power_law_1.2,0.12045439720153808
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,16,4,power_law_1.2,0.12025343894958496
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,16,4,power_law_1.2,0.1533452796936035
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,16,4,power_law_1.2,0.15429504394531252
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,16,4,power_law_1.2,1.6756544494628904
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,16,4,power_law_1.2,0.17368192672729493
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,16,4,power_law_1.2,0.16311424255371093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,16,4,power_law_1.2,0.17336320877075195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,16,4,power_law_1.2,0.1821798324584961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,16,4,power_law_1.2,0.20301055908203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,16,4,power_law_1.2,0.23349119186401368
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,16,4,power_law_1.2,0.28044160842895505
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,16,4,power_law_1.2,2.1709388732910155
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,16,4,power_law_1.2,0.3285567855834961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,16,4,power_law_1.2,0.09597824096679688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,16,4,power_law_1.2,0.44193153381347655
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,32,2,balanced,0.03174527883529663
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,16,4,power_law_1.2,0.552116470336914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,32,2,balanced,0.042707200050354006
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,32,2,balanced,0.060062718391418454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,16,4,power_law_1.2,0.16982784271240234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,16,4,power_law_1.2,0.7251789093017578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,32,2,balanced,0.0728384017944336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,32,2,balanced,0.14560768127441406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,32,2,balanced,0.18858495712280274
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,32,2,balanced,0.19032960891723633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,32,2,balanced,0.19023231506347654
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,32,2,balanced,0.1962278366088867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,32,2,balanced,0.20040191650390624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,32,2,balanced,0.20497152328491214
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,32,2,balanced,0.2383590316772461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,16,4,power_law_1.2,1.328732147216797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,32,2,balanced,0.24316032409667968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,32,2,balanced,0.24732927322387693
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,32,2,balanced,0.255150089263916
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,32,2,balanced,0.2719718360900879
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,32,2,balanced,0.09855487823486328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,32,2,balanced,0.29127935409545896
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,16,4,power_law_1.2,1.7466163635253906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,32,2,balanced,0.37102718353271485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,32,2,balanced,0.19152128219604492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,32,2,balanced,0.4538751983642578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,32,2,balanced,0.0316377592086792
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,32,2,balanced,0.6469888305664062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,32,2,balanced,0.034982399940490724
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,32,2,balanced,0.8558822631835937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,32,2,balanced,0.04283391952514649
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,32,2,balanced,0.05884416103363037
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,32,2,balanced,0.09078016281127929
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,32,2,balanced,0.1456281566619873
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,32,2,balanced,0.3306675338745117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,32,2,balanced,1.1620403289794923
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,32,2,balanced,0.19909503936767578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,16,4,power_law_1.2,0.9280178833007813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,32,2,balanced,0.203636474609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,32,2,balanced,0.20342144012451172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,32,2,balanced,0.20639232635498045
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,32,2,balanced,0.21267839431762697
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,32,2,balanced,0.2109542465209961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,32,2,balanced,1.693219909667969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,32,2,balanced,0.30186111450195313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,32,2,balanced,0.30714752197265627
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,32,2,balanced,0.311596794128418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,32,2,balanced,0.3187968063354492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,32,2,balanced,0.33407360076904297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,32,2,balanced,0.3512307357788086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,32,2,balanced,2.3065843200683593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,32,2,balanced,0.20153343200683596
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,32,2,balanced,0.4264575958251953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,32,2,balanced,0.5038105773925781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,32,2,balanced,0.5736985778808593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,32,2,power_law_1.01,0.03320832014083862
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,32,2,power_law_1.01,0.05731840133666992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,32,2,balanced,0.7287590026855468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,32,2,power_law_1.01,0.06920576095581055
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,32,2,power_law_1.01,0.07881087779998779
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,32,2,balanced,1.0638015747070313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,32,2,balanced,0.387823371887207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,32,2,power_law_1.01,0.10335359573364258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,32,2,power_law_1.01,0.12825471878051758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,32,2,power_law_1.01,0.1369587230682373
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,32,2,power_law_1.01,0.15590911865234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,32,2,power_law_1.01,0.16592767715454101
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,32,2,power_law_1.01,0.17229824066162108
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,32,2,power_law_1.01,0.17559680938720704
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,32,2,power_law_1.01,0.18516992568969726
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,32,2,power_law_1.01,0.214652156829834
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,32,2,power_law_1.01,0.2205683135986328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,32,2,power_law_1.01,0.22393215179443357
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,32,2,balanced,1.9051609802246094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,32,2,power_law_1.01,0.23506048202514646
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,32,2,power_law_1.01,0.2564646339416504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,32,2,power_law_1.01,0.28685440063476564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,32,2,power_law_1.01,0.3350681686401367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,32,2,power_law_1.01,0.3827097702026367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,32,2,power_law_1.01,0.16030847549438476
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,32,2,power_law_1.01,0.48213375091552735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,32,2,balanced,1.3717208862304688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,32,2,power_law_1.01,0.6493926239013672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,32,2,power_law_1.01,0.034982399940490724
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,32,2,power_law_1.01,0.04094207763671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,32,2,power_law_1.01,0.06958720207214356
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,32,2,power_law_1.01,0.05688191890716553
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,32,2,power_law_1.01,0.09397760391235352
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,32,2,power_law_1.01,0.11887359619140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,32,2,power_law_1.01,1.180453109741211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,32,2,power_law_1.01,0.13270400047302247
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,32,2,power_law_1.01,0.15829248428344728
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,32,2,power_law_1.01,0.16417407989501953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,32,2,power_law_1.01,0.1716876792907715
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,32,2,power_law_1.01,0.17559167861938477
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,32,2,power_law_1.01,0.18103296279907227
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,32,2,power_law_1.01,1.7031245422363281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,32,2,power_law_1.01,0.26831487655639646
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,32,2,power_law_1.01,0.2671219253540039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,32,2,power_law_1.01,0.9121804809570312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,32,2,power_law_1.01,0.27176704406738283
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,32,2,power_law_1.01,0.28289535522460935
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,32,2,power_law_1.01,0.30703487396240237
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,32,2,power_law_1.01,0.33535488128662105
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,32,2,power_law_1.01,2.20434814453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,32,2,power_law_1.01,0.38458751678466796
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,32,2,power_law_1.01,0.43597057342529294
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,32,2,power_law_1.01,0.5379673767089843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,32,2,power_law_1.01,0.17639808654785155
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,32,2,power_law_1.2,0.033318400382995605
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,32,2,power_law_1.2,0.057960958480834955
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,32,2,power_law_1.2,0.06654463768005371
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,32,2,power_law_1.2,0.07857279777526856
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,32,2,power_law_1.01,0.8891136169433593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,32,2,power_law_1.2,0.10267647743225097
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,32,2,power_law_1.2,0.12514304161071776
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,32,2,power_law_1.01,1.1140595245361329
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,32,2,power_law_1.2,0.1360896015167236
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,32,2,power_law_1.2,0.15672191619873047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,32,2,power_law_1.2,0.15904640197753905
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,32,2,power_law_1.2,0.16677631378173827
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,32,2,power_law_1.2,0.16845695495605467
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,32,2,power_law_1.01,1.5807513427734374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,32,2,power_law_1.01,0.6536780548095703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,32,2,power_law_1.2,0.18678655624389648
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,32,2,power_law_1.2,0.2160985565185547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,32,2,power_law_1.2,0.21781248092651367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,32,2,power_law_1.2,0.22280319213867186
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,32,2,power_law_1.2,0.23697280883789062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,32,2,power_law_1.2,0.26051328659057615
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,32,2,power_law_1.01,2.0069183349609374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,32,2,power_law_1.2,0.2905254364013672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,32,2,power_law_1.2,0.33931392669677735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,32,2,power_law_1.2,0.3930944061279297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,32,2,power_law_1.2,0.49361919403076177
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,32,2,power_law_1.2,0.034749441146850586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,32,2,power_law_1.2,0.17705343246459962
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,32,2,power_law_1.2,0.6722573089599609
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,32,2,power_law_1.2,0.040862717628479005
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,32,2,power_law_1.2,0.9300556945800782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,32,2,power_law_1.2,0.05497471809387207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,32,2,power_law_1.2,0.06980735778808593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,32,2,power_law_1.2,0.08849535942077637
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,32,2,power_law_1.2,1.19844482421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,32,2,power_law_1.2,0.11760255813598633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,32,2,power_law_1.2,0.1585433578491211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,32,2,power_law_1.2,0.16338815689086914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,32,2,power_law_1.2,1.693395233154297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,32,2,power_law_1.2,0.1694041633605957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,32,2,power_law_1.2,0.17625856399536133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,32,2,power_law_1.2,0.18253183364868164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,32,2,power_law_1.2,0.26516864776611326
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,32,2,power_law_1.2,0.26953983306884766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,32,2,power_law_1.2,0.2706572723388672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,32,2,power_law_1.2,2.246792907714844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,32,2,power_law_1.2,0.12627840042114258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,32,2,power_law_1.2,0.31141248703002933
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,32,2,power_law_1.2,0.3402227020263672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,32,2,power_law_1.2,0.3901900863647461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,32,2,power_law_1.2,0.17067136764526367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,32,2,power_law_1.2,0.4531852722167969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,32,2,power_law_1.2,0.5545228958129883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,1,64,balanced,0.02337536096572876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,32,2,power_law_1.2,0.6924185943603516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,1,64,balanced,0.02310784101486206
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,1,64,balanced,0.02335360050201416
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,32,2,power_law_1.2,0.9149849700927735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,1,64,balanced,0.04133632183074951
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,1,64,balanced,0.04946303844451904
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,1,64,balanced,0.07926271915435791
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,1,64,balanced,0.0791155195236206
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,1,64,balanced,0.07984511852264405
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,1,64,balanced,0.07993599891662598
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,1,64,balanced,0.08093184471130371
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,1,64,balanced,0.08099072456359864
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,1,64,balanced,0.08125823974609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,1,64,balanced,0.08332672119140624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,1,64,balanced,0.08582528114318848
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,32,2,power_law_1.2,1.6032742309570314
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,1,64,balanced,0.08819199562072753
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,1,64,balanced,0.09153663635253906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,1,64,balanced,0.09446656227111816
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,1,64,balanced,0.09642496109008789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,1,64,balanced,0.09102463722229004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,1,64,balanced,0.11786239624023438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,32,2,power_law_1.2,0.28553472518920897
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,1,64,balanced,0.11964672088623048
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,1,64,balanced,0.1629376029968262
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,32,2,power_law_1.2,2.0397914123535155
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,1,64,balanced,0.1941926383972168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,1,64,balanced,0.2756710433959961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,1,64,balanced,0.024999680519104003
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,32,2,power_law_1.2,1.1550438690185547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,1,64,balanced,0.024983038902282713
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,1,64,balanced,0.3302463912963867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,1,64,balanced,0.025491199493408202
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,1,64,balanced,0.035503358840942384
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,1,64,balanced,0.04485504150390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,1,64,balanced,0.043955202102661135
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,1,64,balanced,0.04494336128234863
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,1,64,balanced,0.04490880012512207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,1,64,balanced,0.04706048011779786
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,1,64,balanced,0.046594557762145994
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,1,64,balanced,0.04735104084014893
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,1,64,balanced,0.04907008171081543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,1,64,balanced,0.10109312057495116
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,1,64,balanced,0.0502950382232666
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,1,64,balanced,0.02329983949661255
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,1,64,balanced,0.05713280200958252
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,1,64,balanced,0.050554881095886225
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,1,64,balanced,0.05720448017120361
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,1,64,balanced,0.06020736217498779
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,1,64,balanced,0.06912511825561524
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,1,64,balanced,0.07884543895721435
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,1,64,balanced,0.09785728454589844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,1,64,balanced,0.11658623695373535
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,1,64,balanced,0.025166079998016355
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,1,64,balanced,0.16770944595336915
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,1,64,balanced,0.20183551788330076
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,1,64,power_law_1.01,0.049561600685119625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,1,64,power_law_1.01,0.04943359851837158
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,1,64,power_law_1.01,0.05004672050476074
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,1,64,power_law_1.01,0.07635968208312989
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,1,64,power_law_1.01,0.07865344047546387
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,1,64,balanced,0.04889599800109863
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,1,64,power_law_1.01,0.07922560214996338
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,1,64,power_law_1.01,0.08142463684082031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,1,64,balanced,0.05219583988189698
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,1,64,power_law_1.01,0.08058624267578125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,1,64,power_law_1.01,0.0816972827911377
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,1,64,power_law_1.01,0.08935935974121094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,1,64,power_law_1.01,0.0956595230102539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,1,64,power_law_1.01,0.09553407669067383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,1,64,power_law_1.01,0.09276800155639649
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,1,64,power_law_1.01,0.09251199722290039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,1,64,power_law_1.01,0.10710271835327148
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,1,64,balanced,0.025086719989776612
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,1,64,power_law_1.01,0.1174899196624756
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,1,64,power_law_1.01,0.04967040061950684
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,1,64,power_law_1.01,0.14263296127319336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,1,64,power_law_1.01,0.1702707290649414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,1,64,power_law_1.01,0.2509619140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,1,64,power_law_1.01,0.29178239822387697
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,1,64,power_law_1.01,0.4796403121948242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,1,64,power_law_1.01,0.08582271575927734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,1,64,power_law_1.01,0.03535615921020508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,1,64,power_law_1.01,0.035617280006408694
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,1,64,power_law_1.01,0.035594239234924316
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,1,64,power_law_1.01,0.04322559833526611
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,1,64,power_law_1.01,0.04463359832763672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,1,64,power_law_1.01,1.2213619232177735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,1,64,power_law_1.01,0.04597119808197021
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,1,64,power_law_1.01,0.08832127571105956
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,1,64,power_law_1.01,0.04883840084075928
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,1,64,power_law_1.01,0.04832767963409424
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,1,64,power_law_1.01,0.05144447803497314
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,1,64,power_law_1.01,0.05253888130187988
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,1,64,power_law_1.01,0.9963865661621094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,1,64,power_law_1.01,0.05753983974456787
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,1,64,power_law_1.01,0.05483007907867432
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,1,64,power_law_1.01,0.056364798545837404
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,1,64,power_law_1.01,0.06088704109191895
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,1,64,power_law_1.01,0.06691328048706055
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,1,64,power_law_1.01,0.03512831926345825
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,1,64,power_law_1.01,0.046685438156127926
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,1,64,power_law_1.01,0.10995200157165527
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,1,64,power_law_1.01,0.14906240463256837
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,1,64,power_law_1.01,0.1792844772338867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,1,64,power_law_1.01,0.25125375747680667
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,1,64,power_law_1.01,0.38339584350585937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,1,64,power_law_1.2,0.04950655937194824
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,1,64,power_law_1.01,0.05938047885894775
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,1,64,power_law_1.01,0.5790732955932617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,1,64,power_law_1.2,0.04942592144012451
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,1,64,power_law_1.2,0.07631872177124024
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,1,64,power_law_1.01,0.6741043090820312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,1,64,power_law_1.2,0.07595776081085205
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,1,64,power_law_1.01,0.07808512210845947
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,1,64,power_law_1.2,0.07907968044281007
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,1,64,power_law_1.2,0.08105343818664551
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,1,64,power_law_1.2,0.07936255931854248
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,1,64,power_law_1.2,0.08475008010864257
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,1,64,power_law_1.2,0.08506624221801758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,1,64,power_law_1.2,0.09006848335266113
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,1,64,power_law_1.2,0.09094783782958984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,1,64,power_law_1.2,0.08902400016784669
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,1,64,power_law_1.2,0.092542724609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,1,64,power_law_1.2,0.10272895812988281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,1,64,power_law_1.2,0.04988927841186523
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,1,64,power_law_1.2,0.12064255714416502
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,1,64,power_law_1.2,0.14619135856628418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,1,64,power_law_1.01,0.04664576053619385
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,1,64,power_law_1.2,0.17932416915893554
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,1,64,power_law_1.2,0.07447167873382568
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,1,64,power_law_1.2,0.2528512001037598
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,1,64,power_law_1.2,0.31074687957763675
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,1,64,power_law_1.2,0.4644275283813476
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,1,64,power_law_1.2,0.0349785590171814
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,1,64,power_law_1.2,0.09637760162353516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,1,64,power_law_1.2,0.6171993637084962
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,1,64,power_law_1.2,0.0355622410774231
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,1,64,power_law_1.2,0.03540992021560669
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,1,64,power_law_1.2,0.03188607931137085
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,1,64,power_law_1.2,0.04290944099426269
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,1,64,power_law_1.2,1.0313394927978516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,1,64,power_law_1.2,0.04527616024017334
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,1,64,power_law_1.2,0.04590208053588867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,1,64,power_law_1.2,0.0482700777053833
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,1,64,power_law_1.2,0.04778111934661865
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,1,64,power_law_1.2,1.2965017700195314
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,1,64,power_law_1.2,0.05059584140777588
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,1,64,power_law_1.2,0.051774721145629886
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,1,64,power_law_1.2,0.05506175994873047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,1,64,power_law_1.2,0.05680895805358886
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,1,64,power_law_1.2,0.05454463958740234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,1,64,power_law_1.2,0.05846784114837646
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,1,64,power_law_1.2,0.06812799930572509
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,1,64,power_law_1.2,0.08235136032104493
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,1,64,power_law_1.2,0.10612863540649413
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,1,64,power_law_1.2,0.1300723171234131
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,1,64,power_law_1.2,1.885528259277344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,1,64,power_law_1.2,0.18526208877563477
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,1,64,power_law_1.2,0.24078847885131838
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,1,64,power_law_1.01,0.5486195373535157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,2,32,balanced,0.02117503881454468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,1,64,power_law_1.2,0.33636352539062503
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,2,32,balanced,0.020826880931854246
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,1,64,power_law_1.2,0.43026817321777344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,2,32,balanced,0.021091840267181396
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,1,64,power_law_1.2,0.04753151893615722
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,2,32,balanced,0.023470079898834227
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,2,32,balanced,0.03841919898986816
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,2,32,balanced,0.048102397918701176
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,2,32,balanced,0.07199999809265137
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,2,32,balanced,0.07165823936462402
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,2,32,balanced,0.0740723180770874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,2,32,balanced,0.07400832176208497
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,2,32,balanced,0.07806335926055909
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,2,32,balanced,0.07527040004730225
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,2,32,balanced,0.07904895782470703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,2,32,balanced,0.08055551528930664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,2,32,balanced,0.08370047569274902
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,2,32,balanced,0.08774144172668456
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,2,32,balanced,0.07899903774261474
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,1,64,power_law_1.2,0.9730316925048829
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,2,32,balanced,0.0835161590576172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,2,32,balanced,0.08811136245727538
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,2,32,balanced,0.09626367568969726
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,2,32,balanced,0.10254847526550293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,2,32,balanced,0.1360537624359131
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,2,32,balanced,0.1902566337585449
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,2,32,balanced,0.23027711868286133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,2,32,balanced,0.023046400547027588
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,2,32,balanced,0.32793598175048827
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,2,32,balanced,0.02317568063735962
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,2,32,balanced,0.4035916900634765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,2,32,balanced,0.023249919414520266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,2,32,balanced,0.023953919410705564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,2,32,balanced,0.024145920276641846
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,1,64,power_law_1.2,0.7393241882324219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,2,32,balanced,0.035912959575653075
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,2,32,balanced,0.044107518196105956
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,2,32,balanced,0.04456064224243164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,2,32,balanced,0.04448512077331543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,2,32,balanced,0.046964478492736814
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,2,32,balanced,0.047091197967529294
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,2,32,balanced,0.051672320365905765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,2,32,balanced,0.053263359069824226
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,2,32,balanced,0.05651199817657471
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,2,32,balanced,0.12888832092285157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,2,32,balanced,0.047951359748840336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,2,32,balanced,0.05248256206512451
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,2,32,balanced,0.05555840015411377
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,2,32,balanced,0.06221183776855469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,2,32,balanced,0.06880896091461182
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,2,32,balanced,0.08430975914001465
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,2,32,balanced,0.10002047538757323
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,2,32,balanced,0.13283455848693848
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,2,32,balanced,0.15983615875244142
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,1,64,power_law_1.2,0.05273344039916992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,2,32,balanced,0.23522560119628907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,2,32,power_law_1.01,0.04685056209564209
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,2,32,balanced,0.2873983955383301
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,2,32,power_law_1.01,0.04733439922332764
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,2,32,power_law_1.01,0.046905598640441894
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,2,32,balanced,0.04575615882873535
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,2,32,power_law_1.01,0.04828544139862061
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,2,32,power_law_1.01,0.06830719947814942
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,2,32,power_law_1.01,0.06772223949432374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,2,32,power_law_1.01,0.07082240104675293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,2,32,power_law_1.01,0.07532032012939453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,2,32,power_law_1.01,0.07824255943298339
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,2,32,power_law_1.01,0.08038911819458008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,2,32,power_law_1.01,0.08230655670166016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,2,32,power_law_1.01,0.08643839836120606
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,2,32,power_law_1.01,0.08224767684936524
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,2,32,power_law_1.01,0.08314111709594726
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,2,32,power_law_1.01,0.0946560001373291
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,2,32,power_law_1.01,0.10484479904174804
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,2,32,power_law_1.01,0.13384063720703127
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,2,32,balanced,0.04336383819580078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,2,32,power_law_1.01,0.1627827262878418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,2,32,power_law_1.01,0.18787071228027344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,2,32,power_law_1.01,0.2829017639160156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,2,32,power_law_1.01,0.37634944915771484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,2,32,power_law_1.01,0.07101568222045898
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,2,32,power_law_1.01,0.07489920139312745
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,2,32,power_law_1.01,0.034840319156646725
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,2,32,power_law_1.01,0.4500070571899414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,2,32,power_law_1.01,0.03553663969039917
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,2,32,power_law_1.01,0.03481087923049927
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,2,32,power_law_1.01,0.7339993286132812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,2,32,power_law_1.01,0.036102399826049805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,2,32,power_law_1.01,0.0807859230041504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,2,32,power_law_1.01,0.04128255844116211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,2,32,power_law_1.01,0.0448524808883667
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,2,32,power_law_1.01,0.04453887939453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,2,32,power_law_1.01,0.048404479026794435
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,2,32,power_law_1.01,0.04659327983856201
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,2,32,power_law_1.01,0.9595021057128907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,2,32,power_law_1.01,0.0528652811050415
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,2,32,power_law_1.01,0.056643838882446285
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,2,32,power_law_1.01,0.04925695896148682
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,2,32,power_law_1.01,0.05093247890472412
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,2,32,power_law_1.01,0.05993855953216552
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,2,32,power_law_1.01,0.05177216053009033
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,2,32,power_law_1.01,0.06678016185760498
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,2,32,power_law_1.01,0.08744959831237793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,2,32,power_law_1.01,0.10417792320251465
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,2,32,power_law_1.01,0.13661184310913085
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,2,32,power_law_1.01,0.17374464035034182
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,2,32,power_law_1.01,0.23770496368408206
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,2,32,power_law_1.01,0.04071680068969726
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,2,32,power_law_1.2,0.046973438262939454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,2,32,power_law_1.01,0.28838783264160156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,2,32,power_law_1.2,0.047087359428405764
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,2,32,power_law_1.2,0.046924800872802735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,2,32,power_law_1.2,0.04737023830413818
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,2,32,power_law_1.01,0.5259443283081054
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,2,32,power_law_1.01,0.6015756988525391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,2,32,power_law_1.2,0.06539648056030274
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,2,32,power_law_1.2,0.07048192024230956
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,2,32,power_law_1.2,0.07438208103179932
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,2,32,power_law_1.2,0.0760588788986206
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,2,32,power_law_1.2,0.07739007949829102
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,2,32,power_law_1.2,0.07954815864562989
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,2,32,power_law_1.2,0.08362367630004883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,2,32,power_law_1.2,0.08518912315368651
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,2,32,power_law_1.2,0.0814412784576416
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,2,32,power_law_1.2,0.08209280014038087
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,2,32,power_law_1.2,0.08679936408996583
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,2,32,power_law_1.2,0.1028979206085205
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,2,32,power_law_1.2,0.12165504455566407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,2,32,power_law_1.2,0.06548351764678954
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,2,32,power_law_1.2,0.1540480041503906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,2,32,power_law_1.2,0.1814054489135742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,2,32,power_law_1.2,0.281976318359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,2,32,power_law_1.2,0.07450751781463623
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,2,32,power_law_1.2,0.5163967895507813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,2,32,power_law_1.2,0.6861862182617188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,2,32,power_law_1.2,0.0345472002029419
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,2,32,power_law_1.01,0.05129856109619141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,2,32,power_law_1.2,0.03527679920196533
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,2,32,power_law_1.2,0.03556735992431641
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,2,32,power_law_1.2,0.9770291137695313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,2,32,power_law_1.2,0.04122879981994629
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,2,32,power_law_1.2,0.3184947204589844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,2,32,power_law_1.2,0.04379007816314697
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,2,32,power_law_1.2,0.046383361816406246
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,2,32,power_law_1.2,1.260855026245117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,2,32,power_law_1.2,0.04555520057678223
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,2,32,power_law_1.2,0.04591360092163086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,2,32,power_law_1.2,0.052074241638183597
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,2,32,power_law_1.2,0.05413248062133789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,2,32,power_law_1.2,0.053971199989318853
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,2,32,power_law_1.2,0.05215360164642334
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,2,32,power_law_1.01,0.04415616035461426
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,2,32,power_law_1.2,0.053600001335144046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,2,32,power_law_1.2,0.05467775821685791
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,2,32,power_law_1.2,0.06583168029785155
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,2,32,power_law_1.2,0.08326911926269531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,2,32,power_law_1.2,0.03564160108566285
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,2,32,power_law_1.2,0.11600000381469726
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,2,32,power_law_1.2,0.04057983875274658
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,2,32,power_law_1.2,0.16994432449340818
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,2,32,power_law_1.2,0.1932467269897461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,4,16,balanced,0.0200217604637146
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,2,32,power_law_1.2,0.2846899223327637
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,4,16,balanced,0.020183041095733642
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,2,32,power_law_1.2,0.39865726470947266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,4,16,balanced,0.02143359899520874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,4,16,balanced,0.027023360729217526
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,2,32,power_law_1.2,0.05067008018493653
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,4,16,balanced,0.038754560947418214
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,2,32,power_law_1.2,0.5757132720947266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,4,16,balanced,0.07174143791198731
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,4,16,balanced,0.07202816009521484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,4,16,balanced,0.07618048191070556
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,4,16,balanced,0.08169599533081054
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,4,16,balanced,0.08305791854858399
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,4,16,balanced,0.07624063968658448
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,4,16,balanced,0.07792255878448487
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,4,16,balanced,0.07981184005737305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,4,16,balanced,0.08275967597961426
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,2,32,power_law_1.2,0.09656576156616212
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,4,16,balanced,0.08953856468200684
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,4,16,balanced,0.09650303840637206
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,4,16,balanced,0.11355648040771485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,4,16,balanced,0.1245952033996582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,4,16,balanced,0.15486080169677735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,4,16,balanced,0.17603200912475586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,4,16,balanced,0.2509708786010742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,4,16,balanced,0.31402496337890623
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,4,16,balanced,0.02251136064529419
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,4,16,balanced,0.07522687911987305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,4,16,balanced,0.022383360862731932
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,4,16,balanced,0.4466444778442383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,4,16,balanced,0.07795711994171142
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,4,16,balanced,0.023459839820861816
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,4,16,balanced,0.02672384023666382
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,4,16,balanced,0.5639756774902344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,4,16,balanced,0.03995392084121704
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,4,16,balanced,0.05173632144927979
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,4,16,balanced,0.04946688175201416
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,4,16,balanced,0.050816001892089846
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,4,16,balanced,0.053326082229614255
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,4,16,balanced,0.05548543930053711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,4,16,balanced,0.05609600067138672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,4,16,balanced,0.049305601119995116
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,4,16,balanced,0.05209727764129639
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,4,16,balanced,0.053034238815307624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,4,16,balanced,0.055577597618103027
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,4,16,balanced,0.05975423812866211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,4,16,balanced,0.06756224155426024
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,4,16,balanced,0.08137984275817871
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,4,16,balanced,0.09238911628723144
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,4,16,balanced,0.11998208045959473
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,4,16,balanced,0.05045887947082519
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,4,16,balanced,0.1985139274597168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,2,32,power_law_1.2,0.7682611083984374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,4,16,balanced,0.24434816360473635
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,4,16,balanced,0.36265342712402343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,4,16,power_law_1.01,0.03689599990844726
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,4,16,power_law_1.01,0.047229437828063964
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,4,16,balanced,0.4616371154785156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,4,16,power_law_1.01,0.04786943912506104
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,4,16,power_law_1.01,0.0468287992477417
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,4,16,power_law_1.01,0.06740352153778076
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,4,16,power_law_1.01,0.07047808170318604
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,4,16,balanced,0.14319616317749023
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,4,16,power_law_1.01,0.07316991806030274
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,4,16,power_law_1.01,0.07611775875091553
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,4,16,balanced,0.02273792028427124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,4,16,power_law_1.01,0.07925888061523437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,4,16,power_law_1.01,0.08388735771179198
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,4,16,power_law_1.01,0.07846399784088134
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,4,16,balanced,0.051636481285095216
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,4,16,power_law_1.01,0.0778764820098877
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,4,16,power_law_1.01,0.08042752265930175
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,4,16,power_law_1.01,0.08426239967346191
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,4,16,power_law_1.01,0.09699711799621583
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,4,16,power_law_1.01,0.0483673620223999
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,4,16,power_law_1.01,0.10897279739379884
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,4,16,power_law_1.01,0.126810884475708
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,4,16,power_law_1.01,0.14357760429382324
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,4,16,power_law_1.01,0.18259328842163086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,4,16,power_law_1.01,0.06984576225280761
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,4,16,power_law_1.01,0.2560153579711914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,4,16,power_law_1.01,0.37288959503173824
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,4,16,power_law_1.01,0.02442496061325073
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,4,16,power_law_1.01,0.03759104013442993
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,4,16,power_law_1.01,0.03791743993759155
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,4,16,power_law_1.01,0.46542465209960937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,4,16,power_law_1.01,0.08188544273376465
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,4,16,power_law_1.01,0.03899775981903077
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,4,16,power_law_1.01,0.03877376079559326
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,4,16,power_law_1.01,0.04666368007659912
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,4,16,power_law_1.01,0.048864002227783206
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,4,16,power_law_1.01,0.04959104061126709
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,4,16,power_law_1.01,0.05163136005401612
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,4,16,power_law_1.01,0.0544985580444336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,4,16,power_law_1.01,0.05778816223144532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,4,16,power_law_1.01,0.05751423835754395
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,4,16,power_law_1.01,0.9043609619140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,4,16,power_law_1.01,0.052698879241943365
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,4,16,power_law_1.01,0.052908802032470705
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,4,16,power_law_1.01,0.05460480213165283
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,4,16,power_law_1.01,0.05713151931762696
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,4,16,power_law_1.01,0.0670195198059082
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,4,16,power_law_1.01,0.0767846393585205
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,4,16,power_law_1.01,0.09735424041748046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,4,16,power_law_1.01,0.11743359565734864
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,4,16,power_law_1.01,0.14731648445129394
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,4,16,power_law_1.01,0.16994688034057617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,4,16,power_law_1.01,0.27387008666992185
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,4,16,power_law_1.2,0.036976640224456785
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,4,16,power_law_1.01,0.3412492752075195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,4,16,power_law_1.2,0.047443199157714847
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,4,16,power_law_1.01,0.6672512054443359
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,4,16,power_law_1.2,0.04580095767974853
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,4,16,power_law_1.01,0.0485964822769165
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,4,16,power_law_1.2,0.04826752185821533
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,4,16,power_law_1.01,0.5470080184936523
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,4,16,power_law_1.2,0.050348801612854
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,4,16,power_law_1.2,0.06610943794250487
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,4,16,power_law_1.2,0.07104383945465088
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,4,16,power_law_1.2,0.06886655807495116
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,4,16,power_law_1.2,0.074967041015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,4,16,power_law_1.2,0.0751974391937256
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,4,16,power_law_1.2,0.07889279842376709
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,4,16,power_law_1.2,0.08268927574157715
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,4,16,power_law_1.2,0.08140031814575195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,4,16,power_law_1.2,0.07904767990112305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,4,16,power_law_1.2,0.0850432014465332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,4,16,power_law_1.2,0.10178815841674804
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,4,16,power_law_1.2,0.1159398365020752
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,4,16,power_law_1.2,0.1386342430114746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,4,16,power_law_1.2,0.1593984031677246
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,4,16,power_law_1.2,0.2509286308288574
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,4,16,power_law_1.2,0.31057024002075195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,4,16,power_law_1.2,0.44195201873779294
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,4,16,power_law_1.2,0.024499199390411376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,4,16,power_law_1.01,0.6731455993652344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,4,16,power_law_1.2,0.5291059112548828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,4,16,power_law_1.2,0.036812798976898195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,4,16,power_law_1.2,0.0364518404006958
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,4,16,power_law_1.2,0.08027263641357421
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,4,16,power_law_1.2,0.03810175895690918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,4,16,power_law_1.2,0.03985919952392578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,4,16,power_law_1.2,0.04577792167663574
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,4,16,power_law_1.2,0.049003520011901856
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,4,16,power_law_1.2,0.04891776084899903
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,4,16,power_law_1.2,0.05062911987304688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,4,16,power_law_1.2,0.05121664047241211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,4,16,power_law_1.2,0.05473152160644531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,4,16,power_law_1.2,1.1754227447509766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,4,16,power_law_1.2,0.05806079864501953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,4,16,power_law_1.2,0.051743998527526855
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,4,16,power_law_1.2,0.05389311790466309
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,4,16,power_law_1.2,0.06005375862121583
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,4,16,power_law_1.2,0.05398143768310547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,4,16,power_law_1.2,0.0729088020324707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,4,16,power_law_1.2,0.08466048240661621
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,4,16,power_law_1.2,0.10917632102966308
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,4,16,power_law_1.2,0.1223782444000244
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,4,16,power_law_1.2,0.07705471992492677
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,4,16,power_law_1.2,0.16328191757202148
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,4,16,power_law_1.2,0.8459852600097657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,4,16,power_law_1.2,0.22222335815429686
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,8,8,balanced,0.020207359790802
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,4,16,power_law_1.2,0.2897689628601074
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,8,8,balanced,0.02206207990646362
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,4,16,power_law_1.2,0.43076225280761715
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,8,8,balanced,0.030836479663848875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,4,16,power_law_1.2,0.05884160041809082
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,8,8,balanced,0.040087041854858396
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,8,8,balanced,0.07722367763519286
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,8,8,balanced,0.08033151626586914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,8,8,balanced,0.08182527542114257
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,8,8,balanced,0.08512255668640137
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,8,8,balanced,0.07631999969482421
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,8,8,balanced,0.07826303958892822
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,4,16,power_law_1.2,0.8970751953124999
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,8,8,balanced,0.08073856353759765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,8,8,balanced,0.0826803207397461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,8,8,balanced,0.08513152122497558
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,8,8,balanced,0.09465087890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,8,8,balanced,0.10842880249023437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,8,8,balanced,0.11822336196899415
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,8,8,balanced,0.021419520378112795
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,8,8,balanced,0.14319231986999512
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,8,8,balanced,0.16495744705200194
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,8,8,balanced,0.21651071548461914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,8,8,balanced,0.25742847442626954
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,8,8,balanced,0.05479423999786377
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,8,8,balanced,0.3747174453735352
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,8,8,balanced,0.02302207946777344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,8,8,balanced,0.023394560813903807
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,8,8,balanced,0.4847296142578125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,8,8,balanced,0.023809280395507813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,8,8,balanced,0.02579328060150147
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,8,8,balanced,0.03134720087051392
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,8,8,balanced,0.058090238571167
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,8,8,balanced,0.6994367980957031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,8,8,balanced,0.04912767887115478
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,8,8,balanced,0.06071424007415771
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,8,8,balanced,0.08832639694213867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,8,8,balanced,0.0622489595413208
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,8,8,balanced,0.06337920188903809
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,8,8,balanced,0.05825535774230957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,8,8,balanced,0.06026495933532715
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,8,8,balanced,0.8996659088134766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,8,8,balanced,0.06222976207733154
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,8,8,balanced,0.06343039989471436
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,8,8,balanced,0.06617728233337403
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,8,8,balanced,0.07098112106323243
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,8,8,balanced,0.08475008010864257
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,8,8,balanced,0.11968640327453613
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,8,8,balanced,0.14172800064086916
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,8,8,balanced,0.18852479934692384
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,8,8,balanced,0.23047935485839846
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,8,8,balanced,0.33648769378662113
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,8,8,power_law_1.01,0.022388479709625243
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,8,8,balanced,0.4225779342651367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,8,8,power_law_1.01,0.03853568077087403
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,8,8,power_law_1.01,0.04149248123168946
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,8,8,balanced,0.6255718231201172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,8,8,power_law_1.01,0.044544000625610355
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,8,8,balanced,0.05669248104095459
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,8,8,power_law_1.01,0.046878719329833986
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,8,8,balanced,0.8145715332031249
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,8,8,power_law_1.01,0.050666241645812994
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,8,8,power_law_1.01,0.0743065595626831
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,8,8,power_law_1.01,0.07529088020324706
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,8,8,power_law_1.01,0.07756544113159179
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,8,8,balanced,0.09512831687927245
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,8,8,power_law_1.01,0.0773145580291748
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,8,8,power_law_1.01,0.08130816459655762
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,8,8,power_law_1.01,0.07806335926055909
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,8,8,power_law_1.01,0.08156031608581543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,4,16,power_law_1.2,0.6238105773925782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,8,8,power_law_1.01,0.08001279830932617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,8,8,power_law_1.01,0.08350591659545899
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,8,8,power_law_1.01,0.08871423721313476
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,8,8,power_law_1.01,0.09990400314331055
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,8,8,power_law_1.01,0.11007103919982911
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,8,8,power_law_1.01,0.15734527587890626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,8,8,power_law_1.01,0.21066368103027341
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,8,8,power_law_1.01,0.27578624725341794
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,8,8,power_law_1.01,0.368961296081543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,8,8,power_law_1.01,0.024874238967895507
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,8,8,power_law_1.01,0.027765760421752928
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,8,8,power_law_1.01,0.4894233703613281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,8,8,power_law_1.01,0.03217407941818237
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,8,8,power_law_1.01,0.036287999153137206
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,8,8,power_law_1.01,0.6785062408447265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,8,8,power_law_1.01,0.03980799913406372
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,8,8,power_law_1.01,0.042607359886169434
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,8,8,power_law_1.01,0.077674241065979
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,8,8,power_law_1.01,0.055361280441284175
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,8,8,power_law_1.01,0.05568128108978272
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,8,8,power_law_1.01,0.8924262237548828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,8,8,power_law_1.01,0.05833727836608886
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,8,8,power_law_1.01,0.12970239639282227
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,8,8,power_law_1.01,0.06109312057495118
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,8,8,power_law_1.01,0.054918398857116704
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,8,8,power_law_1.01,0.05901311874389649
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,8,8,power_law_1.01,0.05687935829162598
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,8,8,power_law_1.01,0.06649983882904052
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,8,8,power_law_1.01,0.07952256202697754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,8,8,power_law_1.01,0.10831487655639649
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,8,8,power_law_1.01,0.1339840030670166
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,8,8,power_law_1.01,0.17019392013549806
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,8,8,power_law_1.01,0.2382054328918457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,8,8,power_law_1.01,0.05744128227233887
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,8,8,power_law_1.01,0.06292992115020753
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,8,8,power_law_1.01,0.39887615203857424
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,8,8,power_law_1.01,0.05991551876068115
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,8,8,power_law_1.2,0.022455039024353026
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,8,8,power_law_1.2,0.03842175960540771
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,8,8,power_law_1.01,0.08957568168640137
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,8,8,power_law_1.2,0.0390451192855835
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,8,8,power_law_1.2,0.04325759887695312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,8,8,power_law_1.2,0.048451838493347166
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,8,8,power_law_1.2,0.05402624130249023
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,8,8,power_law_1.2,0.07354623794555663
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,8,8,power_law_1.01,0.76943359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,8,8,power_law_1.2,0.07486080169677735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,8,8,power_law_1.2,0.08171135902404784
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,8,8,power_law_1.01,0.32310142517089846
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,8,8,power_law_1.2,0.0803276824951172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,8,8,power_law_1.2,0.07468031883239747
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,8,8,power_law_1.2,0.07709184169769287
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,8,8,power_law_1.2,0.08203647613525392
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,8,8,power_law_1.2,0.08400768280029297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,8,8,power_law_1.2,0.09017727851867677
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,8,8,power_law_1.2,0.10093184471130372
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,8,8,power_law_1.2,0.11565695762634279
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,8,8,power_law_1.2,0.13940608024597168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,8,8,power_law_1.2,0.16764543533325196
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,8,8,power_law_1.2,0.2156991958618164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,8,8,power_law_1.2,0.07524864196777344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,8,8,power_law_1.2,0.2991423988342285
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,8,8,power_law_1.2,0.42246017456054685
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,8,8,power_law_1.2,0.024890880584716796
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,8,8,power_law_1.2,0.02784640073776245
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,8,8,power_law_1.2,0.5390924835205079
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,8,8,power_law_1.2,0.03142656087875366
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,8,8,power_law_1.2,0.03499264001846313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,8,8,power_law_1.2,0.04085631847381592
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,8,8,power_law_1.2,0.7612518310546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,8,8,power_law_1.2,0.05417471885681152
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,8,8,power_law_1.2,0.054620161056518554
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,8,8,power_law_1.2,0.055544319152832033
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,8,8,power_law_1.2,0.05819263935089112
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,8,8,power_law_1.2,1.0713407897949219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,8,8,power_law_1.2,0.06291584014892579
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,8,8,power_law_1.2,0.05638527870178223
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,8,8,power_law_1.2,0.05787007808685303
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,8,8,power_law_1.2,0.0600767993927002
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,8,8,power_law_1.2,0.062452478408813475
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,8,8,power_law_1.2,0.06783616065979003
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,8,8,power_law_1.2,0.07636735916137696
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,8,8,power_law_1.2,0.08041472434997558
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,8,8,power_law_1.2,0.09179136276245117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,8,8,power_law_1.2,0.11675519943237304
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,8,8,power_law_1.2,0.14083968162536623
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,8,8,power_law_1.2,0.18206079483032228
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,8,8,power_law_1.2,0.03915263891220093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,8,8,power_law_1.2,0.2437388801574707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,8,8,power_law_1.01,0.5794508743286133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,16,4,balanced,0.020095999240875247
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,16,4,balanced,0.02058624029159546
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,8,8,power_law_1.2,0.42559745788574216
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,16,4,balanced,0.022215681076049803
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,8,8,power_law_1.2,0.06013184070587159
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,16,4,balanced,0.058533120155334475
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,16,4,balanced,0.057792000770568855
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,8,8,power_law_1.2,0.6371699142456054
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,16,4,balanced,0.07510272026062012
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,16,4,balanced,0.0761023998260498
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,8,8,power_law_1.2,0.7990937805175781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,16,4,balanced,0.07889279842376709
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,16,4,balanced,0.08070143699645996
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,16,4,balanced,0.08551296234130859
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,16,4,balanced,0.08360960006713866
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,16,4,balanced,0.08855808258056641
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,16,4,balanced,0.09418111801147461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,16,4,balanced,0.1075545597076416
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,16,4,balanced,0.1174899196624756
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,8,8,power_law_1.2,0.3398080062866211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,16,4,balanced,0.1394048023223877
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,16,4,balanced,0.1604083251953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,16,4,balanced,0.04157440185546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,16,4,balanced,0.21264896392822266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,16,4,balanced,0.25396095275878905
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,16,4,balanced,0.37159934997558597
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,16,4,balanced,0.07711616039276123
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,16,4,balanced,0.02221951961517334
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,16,4,balanced,0.022623999118804933
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,16,4,balanced,0.023970561027526854
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,16,4,balanced,0.7094233703613282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,16,4,balanced,0.027372798919677734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,16,4,balanced,0.03352191925048828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,16,4,balanced,0.051151361465454105
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,16,4,balanced,0.05176064014434815
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,16,4,balanced,0.03212671995162964
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,16,4,balanced,0.06540287971496582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,16,4,balanced,0.06611455917358398
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,16,4,balanced,0.06715263843536376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,16,4,balanced,0.059184641838073725
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,16,4,balanced,0.06927103996276855
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,16,4,balanced,0.07112703800201417
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,16,4,balanced,0.07380352020263672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,16,4,balanced,0.07508351802825927
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,16,4,balanced,0.07629695892333985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,16,4,balanced,0.48293247222900393
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,16,4,balanced,0.09371520042419433
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,16,4,balanced,0.10321023941040039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,16,4,balanced,0.12607872009277343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,16,4,balanced,0.1460863971710205
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,16,4,balanced,0.18955392837524415
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,16,4,balanced,0.22980224609375002
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,16,4,balanced,0.33999999999999997
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,16,4,balanced,0.052264962196350094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,16,4,balanced,0.4243724822998047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,16,4,power_law_1.01,0.02136960029602051
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,16,4,power_law_1.01,0.027737600803375246
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,16,4,balanced,0.6305958557128906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,16,4,power_law_1.01,0.036102399826049805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,16,4,power_law_1.01,0.041438717842102055
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,16,4,power_law_1.01,0.04312448024749756
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,16,4,balanced,0.8252684783935547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,16,4,power_law_1.01,0.056942081451416014
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,16,4,power_law_1.01,0.04801919937133789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,16,4,power_law_1.01,0.05685247898101806
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,16,4,power_law_1.01,0.073438720703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,16,4,power_law_1.01,0.0733196783065796
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,16,4,power_law_1.01,0.07445631980895996
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,16,4,power_law_1.01,0.0769484806060791
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,16,4,power_law_1.01,0.07624063968658448
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,16,4,power_law_1.01,0.08032511711120606
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,16,4,power_law_1.01,0.08230400085449219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,16,4,power_law_1.01,0.09214591979980469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,16,4,power_law_1.01,0.10584447860717774
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,16,4,power_law_1.01,0.11727871894836425
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,16,4,power_law_1.01,0.14309632301330566
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,16,4,power_law_1.01,0.16393983840942383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,16,4,power_law_1.01,0.2149785614013672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,16,4,power_law_1.01,0.27557376861572264
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,16,4,balanced,0.08240896224975586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,16,4,power_law_1.01,0.3904153442382813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,16,4,power_law_1.01,0.02343807935714722
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,16,4,power_law_1.01,0.0262719988822937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,16,4,power_law_1.01,0.508714256286621
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,16,4,power_law_1.01,0.02931328058242798
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,16,4,power_law_1.01,0.033093121051788334
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,16,4,power_law_1.01,0.7448076629638672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,16,4,power_law_1.01,0.03840512037277222
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,16,4,power_law_1.01,0.08599424362182617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,16,4,power_law_1.01,0.041000962257385254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,16,4,power_law_1.01,0.0476032018661499
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,16,4,power_law_1.01,0.04772480010986328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,16,4,power_law_1.01,0.0606220817565918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,16,4,power_law_1.01,0.0627788782119751
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,16,4,power_law_1.01,0.9586483001708984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,16,4,power_law_1.01,0.06684544086456298
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,16,4,power_law_1.01,0.06829823970794677
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,16,4,power_law_1.01,0.07047808170318604
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,16,4,power_law_1.01,0.07274112224578858
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,16,4,power_law_1.01,0.09137663841247559
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,16,4,power_law_1.01,0.10324992179870604
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,16,4,power_law_1.01,0.12750080108642578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,16,4,power_law_1.01,0.15384960174560547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,16,4,power_law_1.01,0.19961727142333985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,16,4,power_law_1.01,0.24577407836914061
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,16,4,balanced,0.9204761505126953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,16,4,power_law_1.01,0.36361984252929686
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,16,4,power_law_1.01,0.06256256103515626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,16,4,power_law_1.01,0.06507775783538819
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,16,4,power_law_1.01,0.4625715255737305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,16,4,power_law_1.2,0.02602112054824829
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,16,4,power_law_1.2,0.0348467206954956
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,16,4,power_law_1.01,0.07859839916229247
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,16,4,power_law_1.2,0.039133439064025874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,16,4,power_law_1.2,0.044558081626892086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,16,4,power_law_1.2,0.05506175994873047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,16,4,power_law_1.2,0.07294976234436035
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,16,4,power_law_1.2,0.07350016117095948
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,16,4,power_law_1.01,0.8463654327392579
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,16,4,power_law_1.2,0.0743936014175415
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,16,4,power_law_1.2,0.0762547206878662
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,16,4,power_law_1.2,0.07797376155853272
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,16,4,power_law_1.2,0.0805299186706543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,16,4,power_law_1.2,0.08385663986206054
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,16,4,power_law_1.2,0.08627327919006347
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,16,4,power_law_1.2,0.021427199840545655
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,16,4,power_law_1.2,0.10678912162780763
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,16,4,power_law_1.2,0.1200831985473633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,16,4,power_law_1.01,0.67225341796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,16,4,power_law_1.2,0.16867200851440428
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,16,4,power_law_1.2,0.22148223876953127
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,16,4,power_law_1.2,0.28027008056640623
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,16,4,power_law_1.2,0.40407039642333986
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,16,4,power_law_1.2,0.09215871810913086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,16,4,power_law_1.2,0.532734718322754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,16,4,power_law_1.2,0.0491315221786499
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,16,4,power_law_1.2,0.023379199504852295
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,16,4,power_law_1.2,0.026419200897216798
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,16,4,power_law_1.2,0.7732672119140626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,16,4,power_law_1.2,0.03299583911895752
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,16,4,power_law_1.2,0.037983999252319336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,16,4,power_law_1.2,0.9779264068603515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,16,4,power_law_1.2,0.04701568126678467
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,16,4,power_law_1.2,0.04741888046264649
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,16,4,power_law_1.2,0.061850881576538085
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,16,4,power_law_1.2,0.056906242370605466
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,16,4,power_law_1.2,0.0611891222000122
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,16,4,power_law_1.2,0.06371327877044677
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,16,4,power_law_1.2,0.06899583816528321
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,16,4,power_law_1.2,0.02857856035232544
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,16,4,power_law_1.2,0.07073279857635498
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,16,4,power_law_1.2,0.07306496143341065
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,16,4,power_law_1.2,0.07936384201049804
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,16,4,power_law_1.2,0.09017087936401366
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,16,4,power_law_1.2,0.10685055732727052
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,16,4,power_law_1.2,0.1328089618682861
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,16,4,power_law_1.2,0.15576448440551757
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,16,4,power_law_1.2,0.20605567932128904
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,16,4,power_law_1.2,0.06097536087036133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,16,4,power_law_1.2,0.2641753578186035
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,16,4,power_law_1.2,0.37333503723144534
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,32,2,balanced,0.01997056007385254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,16,4,power_law_1.2,0.14785280227661132
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,32,2,balanced,0.02125312089920044
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,32,2,balanced,0.023527679443359376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,32,2,balanced,0.035289599895477294
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,32,2,balanced,0.04834559917449951
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,32,2,balanced,0.06643328189849854
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,32,2,balanced,0.0677849578857422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,16,4,power_law_1.2,0.6930675506591797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,32,2,balanced,0.0688704013824463
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,16,4,power_law_1.2,0.045263361930847165
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,32,2,balanced,0.0702348804473877
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,32,2,balanced,0.0717568016052246
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,32,2,balanced,0.08669440269470215
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,16,4,power_law_1.2,0.8617049407958983
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,32,2,balanced,0.09229056358337402
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,32,2,balanced,0.09380224227905273
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,32,2,balanced,0.09890048027038574
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,32,2,balanced,0.1090022373199463
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,32,2,balanced,0.11796992301940919
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,32,2,balanced,0.1408345603942871
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,32,2,balanced,0.16099327087402343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,16,4,power_law_1.2,0.4849382400512695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,32,2,balanced,0.2152614402770996
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,32,2,balanced,0.257076473236084
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,32,2,balanced,0.06970623970031739
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,32,2,balanced,0.39196544647216797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,32,2,balanced,0.08774527549743652
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,32,2,balanced,0.500785903930664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,32,2,balanced,0.022083840370178222
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,32,2,balanced,0.08988160133361817
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,32,2,balanced,0.023662080764770506
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,32,2,balanced,0.026734080314636234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,32,2,balanced,0.03168895959854126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,32,2,balanced,0.07156864166259766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,32,2,balanced,0.7463142395019531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,32,2,balanced,0.07265408039093017
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,32,2,balanced,0.07285888195037842
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,32,2,balanced,0.0731712007522583
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,16,4,power_law_1.2,0.06670847892761231
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,32,2,balanced,0.07453951835632325
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,32,2,balanced,0.09821951866149903
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,32,2,balanced,0.0982271957397461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,32,2,balanced,0.10075903892517091
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,32,2,balanced,0.10233728408813476
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,32,2,balanced,0.10379776000976562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,32,2,balanced,0.1088755226135254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,32,2,balanced,0.11858943939208984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,32,2,balanced,0.12671232223510742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,32,2,balanced,0.150383358001709
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,32,2,balanced,0.16729984283447266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,32,2,balanced,0.20743295669555667
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,32,2,balanced,0.24655872344970703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,32,2,balanced,0.38973823547363284
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,32,2,balanced,0.04575104236602783
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,32,2,balanced,0.07189888000488281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,32,2,balanced,0.47003520965576173
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,32,2,power_law_1.01,0.02293247938156128
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,32,2,power_law_1.01,0.034288640022277835
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,32,2,power_law_1.01,0.03858815908432007
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,32,2,balanced,0.6943923187255859
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,32,2,power_law_1.01,0.0470963191986084
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,32,2,balanced,0.9952665710449219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,32,2,power_law_1.01,0.050945281982421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,32,2,power_law_1.01,0.06032639980316162
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,32,2,power_law_1.01,0.06340352058410645
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,32,2,balanced,0.9209932708740235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,32,2,power_law_1.01,0.06871935844421387
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,32,2,power_law_1.01,0.08204416275024415
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,32,2,power_law_1.01,0.08386943817138673
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,32,2,power_law_1.01,0.08712191581726074
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,32,2,power_law_1.01,0.08804224014282228
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,32,2,power_law_1.01,0.09125247955322266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,32,2,power_law_1.01,0.09627519607543945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,32,2,power_law_1.01,0.10878975868225098
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,32,2,power_law_1.01,0.020632319450378418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,32,2,power_law_1.01,0.12096511840820312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,32,2,power_law_1.01,0.14681599617004396
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,32,2,power_law_1.01,0.2202047920227051
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,32,2,power_law_1.01,0.29689344406127927
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,32,2,power_law_1.01,0.06437759876251221
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,32,2,power_law_1.01,0.41381374359130857
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,32,2,power_law_1.01,0.5462041473388671
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,32,2,power_law_1.01,0.023598079681396482
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,32,2,power_law_1.01,0.026054399013519285
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,32,2,power_law_1.01,0.030410239696502684
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,32,2,power_law_1.01,0.7732736206054687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,32,2,power_law_1.01,0.0362060809135437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,32,2,power_law_1.01,0.04417407989501953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,32,2,power_law_1.01,0.06205311775207519
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,32,2,power_law_1.01,0.0494374418258667
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,32,2,power_law_1.01,0.06131072044372558
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,32,2,power_law_1.01,1.0052787017822264
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,32,2,power_law_1.01,0.06276864051818848
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,32,2,power_law_1.01,0.06312575817108154
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,32,2,power_law_1.01,0.06602752208709717
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,32,2,power_law_1.01,0.09011455535888671
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,32,2,power_law_1.01,0.09346176147460937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,32,2,power_law_1.01,0.094136323928833
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,32,2,power_law_1.01,0.09847552299499511
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,32,2,power_law_1.01,0.10378496170043945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,32,2,power_law_1.01,0.1167039966583252
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,32,2,power_law_1.01,0.1370803165435791
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,32,2,power_law_1.01,0.17191808700561523
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,32,2,power_law_1.01,0.17106687545776367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,32,2,power_law_1.01,0.1994598388671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,32,2,power_law_1.01,0.2536652755737305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,32,2,power_law_1.01,0.3042585563659668
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,32,2,power_law_1.01,0.4254822540283203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,32,2,power_law_1.2,0.020730879306793213
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,32,2,power_law_1.2,0.023109118938446045
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,32,2,power_law_1.2,0.034071040153503415
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,32,2,power_law_1.01,0.08925056457519531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,32,2,power_law_1.2,0.03797631978988648
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,32,2,power_law_1.2,0.04587903976440429
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,32,2,power_law_1.01,0.763984603881836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,32,2,power_law_1.2,0.060197119712829594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,32,2,power_law_1.2,0.050335998535156246
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,32,2,power_law_1.2,0.061561598777770996
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,32,2,power_law_1.2,0.06322432041168213
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,32,2,power_law_1.01,0.9947430419921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,32,2,power_law_1.2,0.06891007900238037
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,32,2,power_law_1.2,0.08271871566772461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,32,2,power_law_1.2,0.08714624404907226
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,32,2,power_law_1.2,0.09221887588500977
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,32,2,power_law_1.2,0.08943743705749511
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,32,2,power_law_1.2,0.09789695739746093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,32,2,power_law_1.01,0.05893375873565674
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,32,2,power_law_1.2,0.11160575866699218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,32,2,power_law_1.2,0.12349823951721191
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,32,2,power_law_1.2,0.14960384368896484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,32,2,power_law_1.2,0.17459327697753907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,32,2,power_law_1.01,0.5330828857421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,32,2,power_law_1.2,0.22552959442138673
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,32,2,power_law_1.2,0.3040127944946289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,32,2,power_law_1.2,0.023370239734649658
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,32,2,power_law_1.2,0.43226367950439454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,32,2,power_law_1.2,0.06503424167633057
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,32,2,power_law_1.2,0.02641792058944702
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,32,2,power_law_1.2,0.5556991958618165
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,32,2,power_law_1.2,0.029986560344696045
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,32,2,power_law_1.2,0.03539583921432495
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,32,2,power_law_1.2,0.04291327953338623
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,32,2,power_law_1.2,0.05894400119781494
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,32,2,power_law_1.2,0.06249087810516357
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,32,2,power_law_1.2,1.0361395263671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,32,2,power_law_1.2,0.06405375957489014
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,32,2,power_law_1.2,0.06677887916564942
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,32,2,power_law_1.2,0.08926336288452148
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,32,2,power_law_1.2,0.09116288185119628
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,32,2,power_law_1.2,0.09565055847167969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,32,2,power_law_1.2,0.08439680099487304
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,32,2,power_law_1.2,0.09761792182922363
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,32,2,power_law_1.2,0.10524160385131835
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,32,2,power_law_1.2,0.12252544403076171
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,32,2,power_law_1.2,0.14774016380310057
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,32,2,power_law_1.2,0.1777996826171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,32,2,power_law_1.2,0.2025779151916504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,32,2,power_law_1.2,0.05991424083709716
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,32,2,power_law_1.2,0.2588159942626953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,32,2,power_law_1.2,0.3165324783325195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,1,64,balanced,0.023272960186004638
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,32,2,power_law_1.2,0.09322239875793456
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,1,64,balanced,0.023244800567626952
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,32,2,power_law_1.2,0.5526860809326172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,1,64,balanced,0.0233625602722168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,32,2,power_law_1.2,0.7858995056152344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,1,64,balanced,0.041799678802490234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,32,2,power_law_1.2,0.7763903808593751
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,1,64,balanced,0.06865407943725585
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,1,64,balanced,0.06868991851806641
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,1,64,balanced,0.06974336147308349
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,1,64,balanced,0.0693836784362793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,32,2,power_law_1.2,1.009139175415039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,1,64,balanced,0.0694643211364746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,1,64,balanced,0.07160192012786866
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,1,64,balanced,0.07181695938110352
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,1,64,balanced,0.07199999809265137
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,1,64,balanced,0.07322112083435059
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,1,64,balanced,0.0761625623703003
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,32,2,power_law_1.2,0.430120964050293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,1,64,balanced,0.07982592105865478
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,1,64,balanced,0.08377344131469726
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,1,64,balanced,0.0897548770904541
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,1,64,balanced,0.08359423637390137
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,1,64,balanced,0.08898943901062012
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,1,64,balanced,0.09565183639526367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,1,64,balanced,0.12414079666137696
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,1,64,balanced,0.1415999984741211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,1,64,balanced,0.06922239780426025
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,1,64,balanced,0.18711936950683594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,32,2,power_law_1.2,0.04824960231781006
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,1,64,balanced,0.24203903198242188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,1,64,power_law_1.01,0.04201087951660156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,1,64,power_law_1.01,0.041831679344177246
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,1,64,power_law_1.01,0.06619135856628418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,1,64,power_law_1.01,0.0692416000366211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,1,64,balanced,0.07544960021972656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,1,64,balanced,0.07666175842285157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,1,64,power_law_1.01,0.07043583869934082
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,1,64,power_law_1.01,0.07121664047241211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,1,64,power_law_1.01,0.07362688064575196
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,1,64,power_law_1.01,0.07400447845458984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,1,64,power_law_1.01,0.07550975799560547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,1,64,power_law_1.01,0.0786841583251953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,1,64,power_law_1.01,0.08224639892578126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,1,64,power_law_1.01,0.08692095756530763
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,1,64,power_law_1.01,0.08888319969177247
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,1,64,power_law_1.01,0.08514047622680664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,1,64,power_law_1.01,0.09144576072692871
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,1,64,power_law_1.01,0.10729344367980957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,1,64,power_law_1.01,0.13689087867736816
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,1,64,power_law_1.01,0.16481664657592773
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,1,64,power_law_1.01,0.21890176773071288
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,1,64,power_law_1.01,0.2662092781066895
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,1,64,power_law_1.01,0.06946944236755372
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,1,64,power_law_1.01,0.48535167694091796
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,1,64,power_law_1.2,0.04209152221679687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,1,64,power_law_1.01,0.5645209503173828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,1,64,power_law_1.2,0.04195199966430664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,1,64,power_law_1.2,0.0668236780166626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,1,64,power_law_1.2,0.06946559906005859
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,1,64,power_law_1.2,0.06985472202301025
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,1,64,power_law_1.01,0.8906380462646485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,1,64,power_law_1.2,0.07104640007019043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,1,64,power_law_1.01,0.041818881034851076
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,1,64,power_law_1.2,0.07229568004608154
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,1,64,power_law_1.2,0.07350527763366699
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,1,64,power_law_1.2,0.07703936100006104
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,1,64,power_law_1.2,0.07446271896362304
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,1,64,power_law_1.2,0.0783846378326416
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,1,64,power_law_1.2,0.08128767967224121
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,1,64,power_law_1.2,0.08457471847534179
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,1,64,power_law_1.2,0.08840191841125489
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,1,64,power_law_1.2,0.08743552207946778
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,1,64,power_law_1.2,0.08714879989624023
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,1,64,power_law_1.2,0.11947903633117676
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,1,64,power_law_1.01,0.07738624095916749
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,1,64,power_law_1.2,0.16457344055175782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,1,64,power_law_1.2,0.22008192062377932
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,1,64,power_law_1.2,0.3518860626220703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,1,64,power_law_1.2,0.39865089416503907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,2,32,balanced,0.02006400108337402
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,1,64,power_law_1.2,0.6147647857666015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,1,64,power_law_1.01,1.1218188476562498
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,2,32,balanced,0.020147199630737307
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,2,32,balanced,0.020229120254516602
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,2,32,balanced,0.029960958957672118
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,2,32,balanced,0.051423997879028314
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,1,64,power_law_1.2,0.9729357147216797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,2,32,balanced,0.05979648113250733
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,1,64,power_law_1.2,0.041680641174316406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,2,32,balanced,0.060142078399658204
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,2,32,balanced,0.06055935859680176
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,2,32,balanced,0.06091519832611084
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,2,32,balanced,0.06121984004974366
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,2,32,balanced,0.062334718704223635
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,2,32,balanced,0.06431871891021729
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,2,32,balanced,0.06426368236541749
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,2,32,balanced,0.06772352218627929
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,2,32,balanced,0.0680191993713379
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,2,32,balanced,0.07155456066131591
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,2,32,balanced,0.07767168045043946
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,2,32,balanced,0.07139455795288085
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,2,32,balanced,0.07286911964416504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,2,32,balanced,0.07971072196960449
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,2,32,balanced,0.09093759536743165
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,2,32,balanced,0.09406975746154786
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,2,32,balanced,0.128920316696167
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,2,32,balanced,0.1538700771331787
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,1,64,power_law_1.2,0.1023526382446289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,2,32,balanced,0.2101568031311035
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,2,32,balanced,0.05887487888336181
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,2,32,power_law_1.01,0.03731328010559082
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,2,32,balanced,0.28016128540039065
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,2,32,power_law_1.01,0.03720704078674316
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,2,32,power_law_1.01,0.03793407917022705
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,2,32,power_law_1.01,0.05439104080200195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,2,32,power_law_1.01,0.053939199447631835
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,2,32,power_law_1.01,0.05739776134490967
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,2,32,power_law_1.01,0.05838079929351807
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,2,32,power_law_1.01,0.06002431869506836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,2,32,power_law_1.01,0.06180863857269288
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,2,32,power_law_1.01,0.06261375904083252
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,2,32,power_law_1.01,0.0646233606338501
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,2,32,power_law_1.01,0.0665830421447754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,2,32,power_law_1.01,0.07028480052947998
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,2,32,power_law_1.01,0.07446271896362304
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,2,32,power_law_1.01,0.07743103981018067
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,2,32,power_law_1.01,0.07323904037475586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,2,32,power_law_1.01,0.07952640056610108
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,2,32,power_law_1.01,0.08986495971679688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,2,32,power_law_1.01,0.11440383911132812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,1,64,power_law_1.2,2.135589141845703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,2,32,power_law_1.01,0.14202239990234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,2,32,power_law_1.01,0.20232959747314455
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,2,32,power_law_1.01,0.23216640472412112
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,2,32,power_law_1.01,0.3661824035644531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,2,32,power_law_1.2,0.03749759912490845
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,2,32,power_law_1.01,0.44803199768066404
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,2,32,power_law_1.2,0.03740799903869629
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,2,32,power_law_1.2,0.03757312059402466
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,2,32,power_law_1.01,0.6784703826904297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,2,32,power_law_1.2,0.05495935916900635
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,2,32,power_law_1.01,0.07190400123596191
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,2,32,power_law_1.2,0.05505280017852783
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,2,32,power_law_1.2,0.05758463859558105
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,2,32,power_law_1.2,0.059123201370239256
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,2,32,power_law_1.01,0.928821792602539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,2,32,power_law_1.2,0.060684800148010254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,2,32,power_law_1.2,0.06274687767028808
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,2,32,power_law_1.2,0.06442880153656005
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,2,32,power_law_1.2,0.06844160079956055
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,2,32,power_law_1.2,0.07713280200958253
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,2,32,power_law_1.2,0.07038591861724854
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,2,32,power_law_1.2,0.07420544147491455
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,2,32,power_law_1.2,0.07524352073669434
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,2,32,power_law_1.2,0.08509440422058105
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,2,32,power_law_1.2,0.09926655769348144
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,2,32,power_law_1.2,0.12789119720458983
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,2,32,power_law_1.2,0.1591104030609131
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,2,32,power_law_1.2,0.23685119628906248
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,2,32,power_law_1.2,0.2735590362548828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,4,16,balanced,0.01899392008781433
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,2,32,power_law_1.2,0.4644480133056641
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,2,32,power_law_1.2,0.06271999835968017
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,2,32,power_law_1.2,0.566168327331543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,4,16,balanced,0.01950592041015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,4,16,balanced,0.021962239742279052
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,4,16,balanced,0.030725119113922117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,4,16,balanced,0.05558784008026123
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,4,16,balanced,0.0591871976852417
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,4,16,balanced,0.057543678283691405
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,4,16,balanced,0.05828991889953613
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,4,16,balanced,0.06082816123962402
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,4,16,balanced,0.06162303924560547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,4,16,balanced,0.06280064105987548
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,4,16,balanced,0.06395008087158202
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,4,16,balanced,0.07072512149810792
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,4,16,balanced,0.07427072048187255
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,4,16,balanced,0.06520703792572022
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,1,64,power_law_1.2,1.5780825805664063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,4,16,balanced,0.06733312129974364
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,4,16,balanced,0.07292416095733642
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,4,16,balanced,0.018947839736938477
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,4,16,balanced,0.08354175567626954
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,4,16,balanced,0.0914367961883545
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,4,16,balanced,0.1066368007659912
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,4,16,balanced,0.12165632247924804
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,2,32,power_law_1.2,0.07300992012023926
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,4,16,balanced,0.17052160263061525
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,4,16,balanced,0.20946815490722653
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,4,16,power_law_1.01,0.02908031940460205
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,4,16,balanced,0.2936473655700683
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,4,16,power_law_1.01,0.03680255889892578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,4,16,power_law_1.01,0.036043519973754885
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,4,16,power_law_1.01,0.035018239021301266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,4,16,power_law_1.01,0.05068543910980224
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,4,16,power_law_1.01,0.05566592216491699
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,4,16,power_law_1.01,0.05742464065551758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,4,16,power_law_1.01,0.057731838226318354
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,4,16,power_law_1.01,0.05892352104187012
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,4,16,power_law_1.01,0.06050303936004638
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,4,16,power_law_1.01,0.06266111850738525
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,4,16,power_law_1.01,0.06466815948486329
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,4,16,power_law_1.01,0.0685913610458374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,4,16,power_law_1.01,0.07362304210662843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,4,16,power_law_1.01,0.06841856002807617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,4,16,power_law_1.01,0.06806272029876709
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,4,16,power_law_1.01,0.06814080238342285
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,4,16,power_law_1.01,0.07576704025268555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,4,16,power_law_1.01,0.08352895736694335
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,4,16,power_law_1.01,0.1032870388031006
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,4,16,power_law_1.01,0.11459584236145019
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,4,16,power_law_1.01,0.16106624603271485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,4,16,power_law_1.01,0.1937139129638672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,4,16,power_law_1.01,0.27865215301513674
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,4,16,power_law_1.01,0.3869235229492188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,4,16,power_law_1.2,0.028652799129486085
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,4,16,power_law_1.01,0.4887744140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,4,16,power_law_1.2,0.03478271961212158
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,4,16,balanced,0.3842342376708984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,4,16,power_law_1.2,0.036599040031433105
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,4,16,balanced,0.06584832191467285
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,4,16,power_law_1.01,0.7875545501708985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,4,16,power_law_1.2,0.03650943994522095
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,4,16,power_law_1.2,0.05060480117797852
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,4,16,power_law_1.2,0.055531520843505856
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,4,16,power_law_1.2,0.05400320053100586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,4,16,power_law_1.2,0.05925119876861572
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,4,16,power_law_1.2,0.059137282371520994
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,4,16,power_law_1.2,0.06032639980316162
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,4,16,power_law_1.2,0.06555520057678223
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,4,16,power_law_1.2,0.06951039791107178
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,4,16,power_law_1.2,0.07498879909515381
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,4,16,power_law_1.2,0.06603392124176025
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,4,16,power_law_1.2,0.06909056186676026
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,4,16,power_law_1.2,0.073853440284729
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,4,16,power_law_1.2,0.08033535957336427
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,4,16,power_law_1.2,0.08551296234130859
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,4,16,power_law_1.2,0.11036671638488768
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,2,32,power_law_1.2,1.0193804931640624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,4,16,power_law_1.2,0.14205568313598632
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,2,32,power_law_1.2,1.3618202209472656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,4,16,power_law_1.2,0.1967424011230469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,4,16,power_law_1.2,0.259737606048584
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,8,8,balanced,0.020655360221862793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,8,8,balanced,0.021401600837707518
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,4,16,power_law_1.2,0.3463935852050781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,8,8,balanced,0.022751998901367188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,8,8,balanced,0.03239680051803589
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,8,8,balanced,0.04218111991882324
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,4,16,power_law_1.2,0.06260608196258545
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,8,8,balanced,0.07147007942199707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,8,8,balanced,0.07411327838897705
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,8,8,balanced,0.0761510419845581
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,8,8,balanced,0.07709824085235596
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,8,8,balanced,0.07898880004882812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,8,8,balanced,0.0809830379486084
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,8,8,balanced,0.08586239814758301
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,8,8,balanced,0.08812671661376953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,4,16,power_law_1.2,0.7560076904296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,8,8,balanced,0.0785088014602661
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,8,8,balanced,0.08161791801452636
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,8,8,balanced,0.07990655899047852
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,8,8,balanced,0.0834879970550537
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,4,16,power_law_1.2,0.8805875396728515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,8,8,balanced,0.08977791786193848
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,8,8,balanced,0.09730303764343262
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,8,8,balanced,0.12701312065124512
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,8,8,balanced,0.13952768325805665
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,8,8,balanced,0.18351999282836914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,8,8,balanced,0.21112064361572264
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,8,8,power_law_1.01,0.023376638889312747
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,8,8,balanced,0.3012710380554199
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,8,8,power_law_1.01,0.039256319999694825
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,8,8,power_law_1.01,0.03713279962539673
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,8,8,balanced,0.5408524703979493
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,8,8,power_law_1.01,0.06728576183319092
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,8,8,power_law_1.01,0.07116159915924072
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,8,8,balanced,0.6941439819335937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,4,16,power_law_1.2,0.44610046386718755
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,8,8,power_law_1.01,0.07310848236083985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,8,8,power_law_1.01,0.07385856151580811
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,8,8,power_law_1.01,0.07694975852966308
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,8,8,power_law_1.01,0.08157567977905274
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,8,8,power_law_1.01,0.08360447883605956
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,8,8,power_law_1.01,0.08681471824645996
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,8,8,power_law_1.01,0.07836927890777587
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,8,8,power_law_1.01,0.08002431869506836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,8,8,power_law_1.01,0.08294400215148925
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,8,8,power_law_1.01,0.0864140796661377
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,8,8,power_law_1.01,0.09575424194335938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,8,8,power_law_1.01,0.10626048088073729
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,8,8,power_law_1.01,0.1330291175842285
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,8,8,power_law_1.01,0.15145855903625488
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,8,8,balanced,0.3758015823364258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,8,8,power_law_1.01,0.20599039077758788
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,8,8,power_law_1.01,0.28632703781127933
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,8,8,power_law_1.01,0.386220817565918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,8,8,power_law_1.2,0.023503360748291017
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,8,8,power_law_1.2,0.03467648029327393
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,8,8,power_law_1.01,0.5650790405273438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,8,8,power_law_1.2,0.03775615930557251
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,8,8,power_law_1.2,0.04216320037841797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,8,8,power_law_1.01,0.7092198181152344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,8,8,power_law_1.2,0.06818175792694092
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,8,8,power_law_1.2,0.0716864013671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,8,8,power_law_1.2,0.07345151901245117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,8,8,power_law_1.01,0.9665203094482422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,8,8,power_law_1.01,0.04046720027923584
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,8,8,power_law_1.2,0.08033151626586914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,8,8,power_law_1.01,0.041080322265625004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,8,8,power_law_1.2,0.08311296463012695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,8,8,power_law_1.2,0.08244223594665527
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,8,8,power_law_1.2,0.07923327922821045
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,8,8,power_law_1.2,0.08281472206115723
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,8,8,power_law_1.2,0.08532735824584961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,8,8,power_law_1.2,0.09175295829772949
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,8,8,power_law_1.2,0.09773823738098145
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,8,8,power_law_1.2,0.04193920135498047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,8,8,power_law_1.2,0.14295424461364745
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,8,8,power_law_1.2,0.17301631927490235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,8,8,power_law_1.2,0.23039487838745115
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,8,8,power_law_1.2,0.2870476722717285
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,8,8,power_law_1.2,0.07584383964538574
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,8,8,power_law_1.2,0.07740416049957274
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,16,4,balanced,0.02145535945892334
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,8,8,power_law_1.2,0.4288959884643555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,16,4,balanced,0.022485759258270264
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,16,4,balanced,0.024005119800567624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,16,4,balanced,0.03656960010528564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,16,4,balanced,0.04964863777160645
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,16,4,balanced,0.06128255844116211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,16,4,balanced,0.07942143917083741
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,8,8,power_law_1.2,0.11614975929260254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,16,4,balanced,0.08365695953369141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,16,4,balanced,0.08482560157775879
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,8,8,power_law_1.2,0.5635417556762695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,16,4,balanced,0.09124352455139159
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,16,4,balanced,0.09267840385437012
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,16,4,balanced,0.082357759475708
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,16,4,balanced,0.08598015785217286
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,16,4,balanced,0.08694399833679199
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,16,4,balanced,0.09060352325439454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,16,4,balanced,0.0986303997039795
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,16,4,balanced,0.10650239944458009
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,16,4,balanced,0.14279040336608886
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,16,4,balanced,0.15676416397094725
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,16,4,balanced,0.20923263549804688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,16,4,balanced,0.2394803237915039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,16,4,balanced,0.37040000915527344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,16,4,balanced,0.08141056060791016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,16,4,balanced,0.4558745574951172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,8,8,power_law_1.2,1.010987548828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,16,4,power_law_1.01,0.023119359016418456
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,16,4,power_law_1.01,0.02776063919067383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,16,4,balanced,0.08460415840148926
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,16,4,balanced,0.6680397033691406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,16,4,power_law_1.01,0.04012671947479248
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,16,4,power_law_1.01,0.07122176170349122
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,16,4,power_law_1.01,0.07281536102294922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,16,4,balanced,0.8537677001953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,16,4,power_law_1.01,0.07939839839935303
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,16,4,power_law_1.01,0.08401535987854004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,16,4,power_law_1.01,0.08691455841064452
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,16,4,power_law_1.01,0.08464384078979492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,16,4,power_law_1.01,0.08288895606994628
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,16,4,power_law_1.01,0.08373760223388672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,16,4,power_law_1.01,0.08599935531616211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,16,4,power_law_1.01,0.09149056434631347
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,16,4,power_law_1.01,0.1018611240386963
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,16,4,power_law_1.01,0.0355404806137085
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,16,4,power_law_1.01,0.1404966354370117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,16,4,power_law_1.01,0.16292863845825195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,16,4,power_law_1.01,0.07359104156494141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,16,4,power_law_1.01,0.20991231918334963
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,16,4,power_law_1.01,0.29254655838012694
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,16,4,power_law_1.01,0.04481664180755615
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,16,4,power_law_1.01,0.05558784008026123
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,16,4,power_law_1.01,0.395285758972168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,16,4,power_law_1.2,0.023104000091552734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,16,4,power_law_1.2,0.02928639888763428
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,16,4,power_law_1.2,0.03559936046600341
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,16,4,power_law_1.2,0.040227842330932614
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,16,4,power_law_1.2,0.04465792179107666
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,16,4,power_law_1.01,0.11549440383911133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,16,4,power_law_1.2,0.05586175918579102
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,16,4,power_law_1.01,0.7628928375244141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,16,4,power_law_1.2,0.07030784130096436
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,16,4,power_law_1.2,0.07250944137573243
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,16,4,power_law_1.2,0.0784768009185791
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,16,4,power_law_1.2,0.0844876766204834
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,16,4,power_law_1.01,0.9781670379638673
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,16,4,power_law_1.2,0.08695296287536622
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,16,4,power_law_1.2,0.08411647796630858
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,16,4,power_law_1.2,0.08382207870483398
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,16,4,power_law_1.2,0.08553471565246581
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,16,4,power_law_1.2,0.09347583770751952
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,16,4,power_law_1.2,0.10630271911621095
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,16,4,power_law_1.2,0.12140800476074218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,8,8,power_law_1.2,0.8047974395751953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,16,4,power_law_1.2,0.14900351524353028
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,16,4,power_law_1.01,0.5063923263549805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,16,4,power_law_1.2,0.16902143478393555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,16,4,power_law_1.2,0.2254515266418457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,16,4,power_law_1.2,0.29042175292968747
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,16,4,power_law_1.2,0.4409382247924805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,32,2,balanced,0.020492799282073974
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,16,4,power_law_1.2,0.07528575897216797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,32,2,balanced,0.022183680534362794
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,32,2,balanced,0.04028672218322754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,32,2,balanced,0.057840638160705574
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,32,2,balanced,0.07190783977508544
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,16,4,power_law_1.2,0.08857728004455566
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,32,2,balanced,0.07431424140930176
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,32,2,balanced,0.07529215812683106
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,32,2,balanced,0.08347647666931153
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,32,2,balanced,0.08404992103576661
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,32,2,balanced,0.08689023971557616
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,32,2,balanced,0.08720255851745605
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,32,2,balanced,0.08977151870727539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,32,2,balanced,0.09164799690246582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,32,2,balanced,0.0913203239440918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,32,2,balanced,0.09339008331298829
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,32,2,balanced,0.09590911865234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,32,2,balanced,0.10343040466308592
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,32,2,balanced,0.10936832427978516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,32,2,balanced,0.026179840564727785
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,32,2,balanced,0.1528204822540283
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,32,2,balanced,0.16679168701171876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,32,2,balanced,0.22390655517578123
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,32,2,balanced,0.2520384025573731
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,32,2,balanced,0.39608192443847656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,32,2,balanced,0.48742527008056646
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,32,2,power_law_1.01,0.021267199516296388
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,32,2,power_law_1.01,0.025326080322265625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,32,2,power_law_1.01,0.03488895893096924
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,32,2,power_law_1.01,0.0422976016998291
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,32,2,power_law_1.01,0.048000001907348634
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,32,2,power_law_1.01,0.06299903869628906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,32,2,balanced,0.9292006683349608
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,32,2,power_law_1.01,0.07750144004821777
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,32,2,power_law_1.01,0.07800447940826416
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,32,2,power_law_1.01,0.08112511634826661
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,16,4,power_law_1.2,1.055768356323242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,32,2,power_law_1.01,0.08517248153686523
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,32,2,power_law_1.01,0.08930815696716309
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,32,2,balanced,0.7177970886230469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,32,2,power_law_1.01,0.0940492820739746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,32,2,power_law_1.01,0.06503680229187012
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,32,2,power_law_1.01,0.09928064346313477
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,16,4,power_law_1.2,0.5499942398071289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,32,2,power_law_1.01,0.11376511573791503
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,32,2,power_law_1.01,0.12519935607910154
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,16,4,power_law_1.2,0.8113983917236329
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,32,2,power_law_1.01,0.15152511596679688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,32,2,power_law_1.01,0.17912063598632813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,32,2,power_law_1.01,0.22861055374145506
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,32,2,power_law_1.01,0.08286848068237304
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,32,2,power_law_1.01,0.29677568435668944
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,32,2,power_law_1.01,0.06704512119293213
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,32,2,power_law_1.2,0.021498880386352538
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,32,2,power_law_1.01,0.41192062377929684
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,32,2,power_law_1.01,0.5189452743530273
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,32,2,power_law_1.2,0.024607999324798586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,32,2,power_law_1.01,0.09002367973327637
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,32,2,power_law_1.2,0.035148799419403076
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,32,2,power_law_1.2,0.04152448177337646
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,32,2,power_law_1.2,0.04624000072479248
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,32,2,power_law_1.2,0.06522240161895751
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,32,2,power_law_1.2,0.06715136051177979
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,32,2,power_law_1.2,0.05996799945831299
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,32,2,power_law_1.01,0.7895423889160156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,32,2,power_law_1.2,0.07831552028656005
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,32,2,power_law_1.2,0.08216192245483397
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,32,2,power_law_1.2,0.08382847785949707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,32,2,power_law_1.2,0.08621824264526368
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,32,2,power_law_1.2,0.09201279640197754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,32,2,power_law_1.2,0.09248895645141601
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,32,2,power_law_1.2,0.0947596836090088
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,32,2,power_law_1.01,0.999731216430664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,32,2,power_law_1.2,0.10149760246276855
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,32,2,power_law_1.2,0.11433471679687499
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,32,2,power_law_1.2,0.1286143970489502
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,32,2,power_law_1.2,0.1558835220336914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,32,2,power_law_1.2,0.18386943817138673
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,32,2,power_law_1.2,0.2326131248474121
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,32,2,power_law_1.2,0.314597110748291
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,1,64,balanced,0.05521152019500732
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,1,64,balanced,0.06498816013336181
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,32,2,power_law_1.2,0.42137214660644534
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,1,64,balanced,0.0756556797027588
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,1,64,balanced,0.12095744132995605
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,1,64,balanced,0.1873983955383301
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,1,64,balanced,0.19363584518432617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,1,64,balanced,0.19637760162353515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,1,64,balanced,0.20349567413330077
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,1,64,balanced,0.20416383743286132
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,1,64,balanced,0.21201791763305664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,1,64,balanced,0.20112895965576172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,1,64,balanced,0.21030015945434571
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,1,64,balanced,0.21794944763183594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,1,64,balanced,0.22712064743041993
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,1,64,balanced,0.23080448150634764
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,1,64,balanced,0.24544000625610352
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,1,64,balanced,0.2514112091064453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,1,64,balanced,0.2736179161071777
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,1,64,balanced,0.30216064453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,32,2,power_law_1.2,0.07658880233764649
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,1,64,balanced,0.40967166900634766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,1,64,balanced,0.46609153747558596
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,1,64,balanced,0.6150067138671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,1,64,balanced,0.7080806732177735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,32,2,power_law_1.2,0.5289728164672851
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,1,64,balanced,0.03419136047363282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,32,2,power_law_1.2,1.0469184112548828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,1,64,balanced,0.04349055767059326
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,1,64,balanced,0.04915071964263916
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,1,64,balanced,1.3112486267089845
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,1,64,balanced,0.07255167961120605
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,1,64,balanced,0.10455295562744141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,1,64,balanced,0.10987648010253906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,1,64,balanced,0.11228927612304687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,1,64,balanced,0.11762559890747071
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,1,64,balanced,0.13091456413269043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,1,64,balanced,1.044702682495117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,1,64,balanced,0.12689663887023925
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,32,2,power_law_1.2,0.8269286346435546
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,1,64,balanced,1.9911705017089845
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,1,64,balanced,0.13538432121276855
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,1,64,balanced,0.14167167663574218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,1,64,balanced,0.1457587242126465
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,1,64,balanced,0.155863037109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,1,64,balanced,0.17875200271606445
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,1,64,balanced,0.16745471954345703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,1,64,balanced,0.19235328674316404
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,1,64,balanced,0.24030336380004882
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,1,64,balanced,0.2828889656066894
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,1,64,balanced,2.6199603271484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,1,64,balanced,0.46802558898925783
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,1,64,balanced,0.11864704132080077
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,1,64,balanced,0.12309247970581054
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,1,64,balanced,0.6681472015380859
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,1,64,power_law_1.01,0.13999999999999999
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,1,64,power_law_1.01,0.1481651210784912
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,1,64,balanced,0.8265650939941406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,1,64,power_law_1.01,0.1490278434753418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,1,64,power_law_1.01,0.1510489559173584
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,1,64,power_law_1.01,0.17507328033447264
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,1,64,power_law_1.01,0.17965696334838868
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,1,64,power_law_1.01,0.18403711318969726
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,1,64,balanced,1.2391001892089843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,1,64,power_law_1.01,0.1924006462097168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,1,64,power_law_1.01,0.18831615447998046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,1,64,power_law_1.01,0.20184064865112306
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,1,64,power_law_1.01,0.2142720031738281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,1,64,power_law_1.01,0.23264768600463864
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,1,64,balanced,1.604419860839844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,1,64,power_law_1.01,0.2561625671386719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,1,64,power_law_1.01,0.275598087310791
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,1,64,power_law_1.01,0.29959808349609374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,1,64,power_law_1.01,0.33480831146240236
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,1,64,power_law_1.01,0.426110725402832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,1,64,power_law_1.01,0.1937830352783203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,1,64,balanced,0.38646015167236325
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,1,64,power_law_1.01,0.4849241638183594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,1,64,power_law_1.01,0.6586099243164062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,1,64,power_law_1.01,0.8256140899658202
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,1,64,power_law_1.01,0.09035391807556152
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,1,64,power_law_1.01,1.234228515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,1,64,power_law_1.01,0.09055999755859376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,1,64,power_law_1.01,0.09136384010314942
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,1,64,power_law_1.01,1.4785650634765626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,1,64,power_law_1.01,0.08212608337402344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,1,64,power_law_1.01,0.09280896186828613
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,1,64,power_law_1.01,0.10342911720275878
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,1,64,power_law_1.01,0.1006809616088867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,1,64,power_law_1.01,0.12264448165893556
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,1,64,power_law_1.01,0.11943679809570314
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,1,64,power_law_1.01,0.12269951820373534
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,1,64,power_law_1.01,0.13656319618225096
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,1,64,power_law_1.01,0.14044159889221192
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,1,64,power_law_1.01,0.14580991744995117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,1,64,power_law_1.01,0.16222335815429687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,1,64,power_law_1.01,0.1706879997253418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,1,64,power_law_1.01,2.510419158935547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,1,64,power_law_1.01,0.19130624771118163
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,1,64,power_law_1.01,0.2018764877319336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,1,64,power_law_1.01,0.23760511398315431
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,1,64,power_law_1.01,0.274256649017334
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,1,64,power_law_1.01,0.3779225540161133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,1,64,power_law_1.01,0.4619647979736328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,1,64,power_law_1.01,3.2815142822265626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,1,64,power_law_1.01,0.6492774200439453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,1,64,power_law_1.01,0.7935155487060547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,1,64,power_law_1.2,0.141724157333374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,1,64,power_law_1.2,0.14265727996826172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,1,64,power_law_1.2,0.15445759773254394
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,1,64,power_law_1.01,1.2674559783935546
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,1,64,power_law_1.2,0.15326080322265626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,1,64,power_law_1.2,0.16222848892211914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,1,64,power_law_1.2,0.18706687927246093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,1,64,power_law_1.01,4.866033935546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,1,64,power_law_1.01,1.7374182128906248
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,1,64,power_law_1.2,0.18743040084838866
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,1,64,power_law_1.2,0.2019264030456543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,1,64,power_law_1.2,0.20618623733520508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,1,64,power_law_1.2,0.20692224502563478
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,1,64,power_law_1.2,0.21729408264160158
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,1,64,power_law_1.2,0.227512321472168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,1,64,power_law_1.2,0.2758515167236328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,1,64,power_law_1.01,2.707796630859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,1,64,power_law_1.2,0.1774732780456543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,1,64,power_law_1.2,0.29115264892578124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,1,64,power_law_1.2,0.3152448081970215
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,1,64,power_law_1.2,0.3813657760620117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,1,64,power_law_1.01,6.710654907226562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,1,64,power_law_1.2,0.4815155029296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,1,64,power_law_1.2,0.5602022552490235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,1,64,power_law_1.01,3.673456726074219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,1,64,power_law_1.2,1.0520665740966797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,1,64,power_law_1.2,1.5181977844238281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,1,64,power_law_1.2,0.7524467468261719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,1,64,power_law_1.2,0.09058943748474121
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,1,64,power_law_1.2,0.09059712409973145
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,1,64,power_law_1.2,0.09156096458435059
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,1,64,power_law_1.2,2.045518035888672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,1,64,power_law_1.2,0.0920358371734619
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,1,64,power_law_1.2,0.09296768188476563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,1,64,power_law_1.2,0.10428031921386718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,1,64,power_law_1.2,0.10360192298889159
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,1,64,power_law_1.2,0.12130047798156737
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,1,64,power_law_1.2,0.12989824295043945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,1,64,power_law_1.2,0.12636544227600097
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,1,64,power_law_1.2,0.13530367851257324
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,1,64,power_law_1.2,0.14231679916381837
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,1,64,power_law_1.2,3.033807373046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,1,64,power_law_1.2,0.16524415969848633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,1,64,power_law_1.2,0.18575103759765627
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,1,64,power_law_1.2,0.18731775283813476
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,1,64,power_law_1.2,0.264901123046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,1,64,power_law_1.2,0.20364160537719728
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,1,64,power_law_1.2,0.32686206817626956
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,1,64,power_law_1.2,0.44163070678710936
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,1,64,power_law_1.2,0.5262323379516601
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,1,64,power_law_1.2,0.14519807815551758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,1,64,power_law_1.2,4.468275146484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,1,64,power_law_1.2,0.8075929260253906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,2,32,balanced,0.0381273603439331
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,1,64,power_law_1.2,1.14382080078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,2,32,balanced,0.05387008190155029
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,2,32,balanced,0.06394624233245849
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,2,32,balanced,0.11589887619018555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,2,32,balanced,0.16416255950927733
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,2,32,balanced,0.19376768112182616
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,2,32,balanced,0.21547903060913084
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,1,64,power_law_1.2,1.6832652282714844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,2,32,balanced,0.20220287322998046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,2,32,balanced,0.21610240936279296
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,2,32,balanced,0.2172812843322754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,2,32,balanced,0.22215295791625977
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,2,32,balanced,0.23615488052368164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,1,64,power_law_1.2,2.3352333068847657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,2,32,balanced,0.24103296279907224
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,2,32,balanced,0.2588607978820801
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,2,32,balanced,0.20013696670532227
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,2,32,balanced,0.2471999931335449
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,2,32,balanced,0.26347263336181637
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,2,32,balanced,0.3217049789428711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,2,32,balanced,0.34955009460449216
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,2,32,balanced,0.47957630157470704
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,2,32,balanced,0.5482662582397462
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,1,64,power_law_1.2,7.034685668945312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,2,32,balanced,0.7265280151367188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,2,32,balanced,0.8654080200195313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,2,32,balanced,0.028015360832214353
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,2,32,balanced,0.03632128000259399
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,2,32,balanced,1.2849600219726562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,2,32,balanced,0.04555776119232178
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,1,64,power_law_1.2,8.7700341796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,2,32,balanced,0.073056001663208
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,2,32,balanced,0.0990668773651123
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,2,32,balanced,1.6264320373535157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,2,32,balanced,0.13164416313171387
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,2,32,balanced,0.12271743774414064
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,2,32,balanced,0.1217958354949951
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,2,32,balanced,0.12274304389953614
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,2,32,balanced,0.12923647880554198
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,2,32,balanced,0.13720704078674317
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,2,32,balanced,0.14978943824768068
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,2,32,balanced,0.1620992088317871
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,1,64,power_law_1.2,3.639578857421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,2,32,balanced,0.17938688278198242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,2,32,balanced,0.1647577667236328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,2,32,balanced,0.1791654396057129
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,2,32,balanced,2.4567858886718748
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,2,32,balanced,0.24399744033813478
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,2,32,balanced,0.30647680282592776
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,2,32,balanced,0.3683686447143555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,2,32,balanced,0.5167308807373047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,2,32,balanced,0.1343513584136963
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,2,32,balanced,0.649201889038086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,2,32,balanced,3.2378393554687497
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,1,64,power_law_1.2,4.413630676269531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,2,32,balanced,0.21049983978271483
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,2,32,power_law_1.01,0.15452544212341307
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,2,32,balanced,1.186297607421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,2,32,power_law_1.01,0.15883904457092285
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,2,32,power_law_1.01,0.16835968017578123
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,2,32,power_law_1.01,0.15986432075500487
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,2,32,power_law_1.01,0.13697792053222657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,2,32,power_law_1.01,0.17882368087768555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,2,32,power_law_1.01,0.181583366394043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,2,32,power_law_1.01,0.1745484733581543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,2,32,balanced,0.9266738891601562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,2,32,power_law_1.01,0.17923967361450194
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,2,32,power_law_1.01,0.18383743286132812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,2,32,power_law_1.01,0.1933247947692871
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,2,32,power_law_1.01,0.21121152877807617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,2,32,power_law_1.01,0.2383820724487305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,2,32,power_law_1.01,0.24230911254882814
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,2,32,power_law_1.01,0.2665484809875488
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,2,32,power_law_1.01,0.2754009628295898
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,2,32,balanced,2.2976448059082033
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,2,32,power_law_1.01,0.3036172866821289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,2,32,power_law_1.01,0.35088897705078126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,2,32,power_law_1.01,0.4181235122680664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,2,32,balanced,1.758033905029297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,2,32,power_law_1.01,0.5534873580932617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,2,32,power_law_1.01,0.7032614135742188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,2,32,power_law_1.01,0.09297663688659667
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,2,32,power_law_1.01,1.0094143676757814
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,2,32,power_law_1.01,0.09270912170410156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,2,32,power_law_1.01,1.36060546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,2,32,power_law_1.01,0.09389183998107911
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,2,32,power_law_1.01,0.09582847595214844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,2,32,power_law_1.01,0.08434687614440918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,2,32,power_law_1.01,0.12149120330810548
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,2,32,power_law_1.01,0.1073305606842041
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,2,32,power_law_1.01,0.12320639610290526
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,2,32,power_law_1.01,1.7648268127441404
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,2,32,power_law_1.01,0.12547455787658693
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,2,32,power_law_1.01,0.13224831581115723
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,2,32,power_law_1.01,0.13669504165649413
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,2,32,power_law_1.01,0.14269439697265623
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,2,32,power_law_1.01,0.17133312225341796
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,2,32,power_law_1.01,0.16821887969970703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,2,32,power_law_1.01,2.412115173339844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,2,32,power_law_1.01,0.18607744216918945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,2,32,power_law_1.01,0.21796863555908202
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,2,32,power_law_1.01,0.26184831619262694
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,2,32,power_law_1.01,0.3556710433959961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,2,32,power_law_1.01,0.455230712890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,2,32,power_law_1.01,0.631852798461914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,2,32,power_law_1.01,0.8212620544433594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,2,32,power_law_1.01,3.8067379760742184
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,2,32,power_law_1.01,0.12965888023376465
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,2,32,power_law_1.2,0.15352191925048828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,2,32,power_law_1.01,1.194672622680664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,2,32,power_law_1.01,0.1747315216064453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,2,32,power_law_1.2,0.1553536033630371
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,2,32,power_law_1.01,1.5385049438476561
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,2,32,power_law_1.2,0.16360960006713868
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,2,32,power_law_1.2,0.14105728149414062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,2,32,power_law_1.01,5.06229248046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,2,32,power_law_1.2,0.17024639129638672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,2,32,power_law_1.2,0.1807539176940918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,2,32,power_law_1.2,0.19975168228149415
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,2,32,power_law_1.2,0.17491840362548827
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,2,32,power_law_1.2,0.1978175926208496
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,2,32,power_law_1.2,0.1671014404296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,2,32,power_law_1.2,0.21192447662353517
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,2,32,power_law_1.2,0.26786304473876954
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,2,32,power_law_1.01,2.349866180419922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,2,32,power_law_1.2,0.2784102439880371
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,2,32,power_law_1.2,0.2853670310974121
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,2,32,power_law_1.2,0.3126924705505371
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,2,32,power_law_1.2,0.40840705871582034
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,2,32,power_law_1.2,0.47113471984863275
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,2,32,power_law_1.2,0.605172462463379
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,2,32,power_law_1.2,0.19670911788940432
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,2,32,power_law_1.2,0.2262451171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,2,32,power_law_1.2,0.8244326019287109
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,2,32,power_law_1.01,3.3379968261718753
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,2,32,power_law_1.2,0.09314816474914552
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,2,32,power_law_1.2,1.171377944946289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,2,32,power_law_1.2,0.10082176208496094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,2,32,power_law_1.2,0.09414400100708008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,2,32,power_law_1.2,0.09574399948120117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,2,32,power_law_1.2,1.4924850463867188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,2,32,power_law_1.2,0.08887807846069336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,2,32,power_law_1.2,0.12337023735046386
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,2,32,power_law_1.2,0.12126591682434082
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,2,32,power_law_1.2,0.1239692783355713
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,2,32,power_law_1.2,0.12908800125122072
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,2,32,power_law_1.2,0.13251968383789064
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,2,32,power_law_1.2,2.206149139404297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,2,32,power_law_1.2,0.14171775817871093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,2,32,power_law_1.2,0.1535257625579834
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,2,32,power_law_1.2,0.1778816032409668
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,2,32,power_law_1.2,0.16997631072998048
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,2,32,power_law_1.2,0.17077632904052736
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,2,32,power_law_1.2,0.19220224380493164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,2,32,power_law_1.2,0.11310336112976074
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,2,32,power_law_1.2,0.2382348823547363
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,2,32,power_law_1.2,2.9191104125976564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,2,32,power_law_1.2,0.28935680389404295
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,2,32,power_law_1.2,0.4100019073486328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,2,32,power_law_1.2,0.49668735504150396
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,2,32,power_law_1.2,0.6666738891601562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,2,32,power_law_1.2,0.9202252960205077
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,4,16,balanced,0.033532159328460695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,2,32,power_law_1.2,1.3638592529296876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,4,16,balanced,0.047735037803649905
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,2,32,power_law_1.2,4.65462646484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,4,16,balanced,0.05988224029541016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,4,16,balanced,0.14427392005920409
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,4,16,balanced,0.17152896881103516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,4,16,balanced,0.19729791641235353
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,2,32,power_law_1.2,1.8178918457031252
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,4,16,balanced,0.20474367141723632
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,4,16,balanced,0.2133452796936035
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,4,16,balanced,0.21443328857421876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,4,16,balanced,0.22347007751464845
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,4,16,balanced,0.2328294372558594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,4,16,balanced,0.23771520614624025
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,4,16,balanced,0.25848447799682617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,4,16,balanced,0.24760576248168947
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,4,16,balanced,0.0895910358428955
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,4,16,balanced,0.25757823944091796
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,2,32,power_law_1.2,2.742577819824219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,4,16,balanced,0.3356787109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,4,16,balanced,0.38517505645751954
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,4,16,balanced,0.5341081619262695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,4,16,balanced,0.6133452987670898
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,4,16,balanced,0.836789779663086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,2,32,power_law_1.2,3.542945251464844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,4,16,balanced,1.0104217529296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,4,16,balanced,0.027317759990692136
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,4,16,balanced,0.030135040283203123
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,2,32,power_law_1.2,6.677485961914063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,4,16,balanced,0.04310656070709228
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,4,16,balanced,0.06087552070617676
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,4,16,balanced,1.501651153564453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,4,16,balanced,0.10154239654541015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,4,16,balanced,0.11096575736999512
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,4,16,balanced,0.1213798427581787
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,4,16,balanced,0.12192895889282226
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,4,16,balanced,0.13231231689453124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,4,16,balanced,0.14089216232299803
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,4,16,balanced,0.1564902400970459
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,4,16,balanced,1.9209907531738282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,4,16,balanced,0.27941759109497066
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,4,16,balanced,0.1769536018371582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,4,16,balanced,0.18185344696044922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,4,16,balanced,0.19641855239868164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,4,16,balanced,0.24106239318847655
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,4,16,balanced,0.28413312911987304
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,4,16,balanced,0.3615359878540039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,4,16,balanced,2.852179260253906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,4,16,balanced,0.13459456443786622
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,4,16,balanced,0.4396940612792969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,4,16,balanced,0.1674777603149414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,4,16,balanced,0.16487039566040038
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,4,16,balanced,0.6401484680175782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,4,16,balanced,0.797870101928711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,4,16,power_law_1.01,0.10847104072570801
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,4,16,power_law_1.01,0.1465344047546387
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,4,16,power_law_1.01,0.13963264465332031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,4,16,balanced,3.724029541015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,4,16,balanced,1.1838655853271485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,4,16,power_law_1.01,0.15951231956481932
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,4,16,power_law_1.01,0.16067712783813476
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,4,16,power_law_1.01,0.19030143737792968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,4,16,power_law_1.01,0.1796019172668457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,4,16,power_law_1.01,0.18688127517700195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,4,16,power_law_1.01,0.18946815490722657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,4,16,power_law_1.01,0.1953638458251953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,4,16,power_law_1.01,0.21000959396362306
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,4,16,power_law_1.01,0.22551424026489258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,4,16,power_law_1.01,0.23974271774291994
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,4,16,balanced,2.248076782226563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,4,16,power_law_1.01,0.2438528060913086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,4,16,power_law_1.01,0.1548262405395508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,4,16,power_law_1.01,0.25078271865844726
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,4,16,power_law_1.01,0.28981632232666016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,4,16,balanced,1.5042022705078124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,4,16,power_law_1.01,0.3563750457763672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,4,16,power_law_1.01,0.4271500778198242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,4,16,power_law_1.01,0.5347737503051758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,4,16,power_law_1.01,0.6429440307617187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,4,16,balanced,2.9340493774414065
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,4,16,power_law_1.01,0.922260513305664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,4,16,power_law_1.01,0.07002240180969238
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,4,16,power_law_1.01,0.10354816436767578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,4,16,power_law_1.01,1.1399155426025391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,4,16,power_law_1.01,0.09629952430725097
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,4,16,power_law_1.01,0.09889920234680176
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,4,16,power_law_1.01,0.10049535751342772
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,4,16,power_law_1.01,0.1065676784515381
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,4,16,power_law_1.01,0.1280998420715332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,4,16,power_law_1.01,0.12407936096191405
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,4,16,power_law_1.01,1.7249574279785158
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,4,16,power_law_1.01,0.13441280364990232
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,4,16,power_law_1.01,0.14268927574157714
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,4,16,power_law_1.01,0.15521408081054688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,4,16,power_law_1.01,2.1956005859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,4,16,power_law_1.01,0.15574399948120116
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,4,16,power_law_1.01,0.16884992599487306
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,4,16,power_law_1.01,0.11770367622375488
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,4,16,power_law_1.01,0.19180288314819335
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,4,16,power_law_1.01,0.12706303596496582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,4,16,power_law_1.01,0.23052671432495114
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,4,16,power_law_1.01,0.2652620887756348
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,4,16,power_law_1.01,0.3628569412231445
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,4,16,power_law_1.01,3.2666149902343746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,4,16,power_law_1.01,0.4442700958251953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,4,16,power_law_1.01,0.17286399841308592
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,4,16,power_law_1.01,0.6341567993164062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,4,16,power_law_1.01,0.7917478179931641
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,4,16,power_law_1.2,0.1090329647064209
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,4,16,power_law_1.01,4.378201599121094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,4,16,power_law_1.2,0.1300556755065918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,4,16,power_law_1.2,0.1434662437438965
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,4,16,power_law_1.01,1.1997657775878907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,4,16,power_law_1.2,0.1676288032531738
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,4,16,power_law_1.2,0.15779711723327636
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,4,16,power_law_1.01,1.5123558044433594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,4,16,power_law_1.2,0.1796518325805664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,4,16,power_law_1.2,0.18332672119140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,4,16,power_law_1.2,0.18957183837890626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,4,16,power_law_1.2,0.21199104309082034
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,4,16,power_law_1.2,0.20162559509277345
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,4,16,power_law_1.2,0.22647424697875979
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,4,16,power_law_1.2,0.250833911895752
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,4,16,power_law_1.2,0.2671232032775879
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,4,16,power_law_1.2,0.2667763137817383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,4,16,power_law_1.2,0.3084671974182129
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,4,16,power_law_1.2,0.18082176208496095
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,4,16,power_law_1.2,0.39086208343505857
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,4,16,power_law_1.2,0.4492083358764648
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,4,16,power_law_1.2,0.5600153732299804
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,4,16,power_law_1.01,2.9349658203124998
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,4,16,power_law_1.2,0.1544371223449707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,4,16,power_law_1.2,0.7160063934326172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,4,16,power_law_1.01,2.315459899902344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,4,16,power_law_1.2,0.07065599918365478
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,4,16,power_law_1.2,0.09666175842285156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,4,16,power_law_1.2,1.3332786560058594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,4,16,power_law_1.2,0.09559295654296876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,4,16,power_law_1.2,0.09846783638000488
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,4,16,power_law_1.2,0.10193023681640626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,4,16,power_law_1.2,0.12395135879516601
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,4,16,power_law_1.2,1.930164489746094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,4,16,power_law_1.2,1.0441970825195312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,4,16,power_law_1.2,0.12365056037902833
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,4,16,power_law_1.2,0.1251097583770752
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,4,16,power_law_1.2,0.14623744010925294
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,4,16,power_law_1.2,2.450556182861328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,4,16,power_law_1.2,0.15724800109863282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,4,16,power_law_1.2,0.09957632064819336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,4,16,power_law_1.2,0.17825536727905272
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,4,16,power_law_1.2,0.16129791259765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,4,16,power_law_1.2,0.16793855667114257
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,4,16,power_law_1.2,0.11742079734802247
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,4,16,power_law_1.2,0.1930905532836914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,4,16,power_law_1.2,0.2409689521789551
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,4,16,power_law_1.2,0.27650047302246095
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,4,16,power_law_1.2,0.13160575866699217
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,4,16,power_law_1.2,0.38328575134277343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,4,16,power_law_1.2,3.5847949218750004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,4,16,power_law_1.2,0.47786113739013675
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,8,8,balanced,0.026484479904174806
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,4,16,power_law_1.2,0.6660915374755859
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,8,8,balanced,0.044199681282043456
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,8,8,balanced,0.054408960342407234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,8,8,balanced,0.07812352180480957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,8,8,balanced,0.11820544242858885
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,8,8,balanced,0.1621811294555664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,8,8,balanced,0.17435264587402344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,4,16,power_law_1.2,1.24302978515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,8,8,balanced,0.1781043243408203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,8,8,balanced,0.21014783859252928
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,8,8,balanced,0.2090572738647461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,4,16,power_law_1.2,1.611094970703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,8,8,balanced,0.2348159980773926
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,4,16,power_law_1.2,0.8661119842529297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,8,8,balanced,0.23791872024536134
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,8,8,balanced,0.2528819274902344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,8,8,balanced,0.23999103546142578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,8,8,balanced,0.2510041618347168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,8,8,balanced,0.2731711959838867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,8,8,balanced,0.31739776611328124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,4,16,power_law_1.2,2.4303065490722657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,8,8,balanced,0.36615169525146485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,8,8,balanced,0.21953664779663087
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,8,8,balanced,0.6022335815429687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,8,8,balanced,0.8385779571533203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,8,8,balanced,1.0081919860839843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,4,16,power_law_1.2,5.071084899902344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,8,8,balanced,0.02619647979736328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,8,8,balanced,0.03075455904006958
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,8,8,balanced,0.04083583831787109
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,8,8,balanced,1.5138983154296874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,8,8,balanced,0.058283519744873044
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,8,8,balanced,0.08982144355773926
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,8,8,balanced,0.11710847854614256
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,8,8,balanced,1.9200320434570313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,8,8,balanced,0.11989503860473634
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,8,8,balanced,0.1467852783203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,8,8,balanced,0.14749183654785156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,8,8,balanced,0.5225740814208985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,8,8,balanced,0.15158528327941895
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,8,8,balanced,0.16264320373535157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,4,16,power_law_1.2,3.252578430175781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,8,8,balanced,0.16899328231811522
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,8,8,balanced,0.1862387275695801
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,8,8,balanced,2.850611267089844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,8,8,balanced,0.10987135887145996
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,8,8,balanced,0.19860095977783204
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,8,8,balanced,0.18886655807495117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,8,8,balanced,0.23574911117553712
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,8,8,balanced,0.2752115249633789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,8,8,balanced,0.34979583740234377
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,8,8,balanced,0.4199270248413086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,8,8,balanced,0.6254604721069337
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,8,8,balanced,0.7708377838134766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,8,8,balanced,0.17030399322509765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,8,8,balanced,3.7062619018554686
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,8,8,power_law_1.01,0.06888959884643556
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,8,8,power_law_1.01,0.09622271537780762
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,8,8,power_law_1.01,0.10641792297363281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,8,8,balanced,1.142513885498047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,8,8,power_law_1.01,0.12057472229003907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,8,8,power_law_1.01,0.13916031837463377
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,8,8,power_law_1.01,0.1579417610168457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,8,8,power_law_1.01,0.15033727645874023
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,8,8,balanced,1.433807373046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,8,8,power_law_1.01,0.16557056427001954
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,8,8,power_law_1.01,0.1927065658569336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,8,8,power_law_1.01,0.21890687942504883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,8,8,power_law_1.01,0.2169011116027832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,8,8,power_law_1.01,0.24146047592163086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,8,8,power_law_1.01,0.24200319290161132
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,8,8,balanced,2.152313537597656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,8,8,power_law_1.01,0.2840409660339355
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,8,8,power_law_1.01,0.335731201171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,8,8,power_law_1.01,0.42953342437744146
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,8,8,power_law_1.01,0.5344230270385741
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,8,8,power_law_1.01,0.6354560089111329
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,8,8,balanced,2.797213439941406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,8,8,power_law_1.01,0.8960857391357422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,8,8,power_law_1.01,0.1904524803161621
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,8,8,power_law_1.01,0.2002943992614746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,8,8,power_law_1.01,1.1323967742919923
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,8,8,power_law_1.01,0.2376063919067383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,8,8,power_law_1.01,0.04877312183380127
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,8,8,power_law_1.01,0.07517312049865724
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,8,8,power_law_1.01,0.06665472030639649
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,8,8,power_law_1.01,0.08694911956787109
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,8,8,power_law_1.01,1.6511053466796874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,8,8,power_law_1.01,0.11035136222839355
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,8,8,power_law_1.01,0.10231552124023438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,8,8,power_law_1.01,0.11390080451965331
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,8,8,power_law_1.01,0.13903360366821288
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,8,8,power_law_1.01,0.14467583656311037
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,8,8,power_law_1.01,0.15499775886535644
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,8,8,power_law_1.01,2.198673858642578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,8,8,power_law_1.01,0.16203264236450193
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,8,8,power_law_1.01,0.1806822395324707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,8,8,power_law_1.01,0.16657663345336915
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,8,8,power_law_1.01,0.17284736633300782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,8,8,power_law_1.01,0.19775104522705078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,8,8,power_law_1.01,0.09414912223815917
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,8,8,power_law_1.01,0.23793792724609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,8,8,power_law_1.01,0.2749350357055664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,8,8,power_law_1.01,0.37800960540771483
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,8,8,power_law_1.01,0.14175359725952147
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,8,8,power_law_1.01,0.4653785705566406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,8,8,power_law_1.01,0.6223078536987304
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,8,8,power_law_1.01,0.8086489868164062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,8,8,power_law_1.2,0.06902656078338623
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,8,8,power_law_1.2,0.08157567977905274
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,8,8,power_law_1.01,1.1524787139892578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,8,8,power_law_1.01,4.04034423828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,8,8,power_law_1.2,0.11953920364379882
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,8,8,power_law_1.01,1.5256422424316407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,8,8,power_law_1.2,0.13209728240966795
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,8,8,power_law_1.2,0.16745855331420897
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,8,8,power_law_1.2,0.14956159591674806
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,8,8,power_law_1.01,3.0014669799804685
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,8,8,power_law_1.2,0.15704575538635254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,8,8,power_law_1.2,0.18648704528808593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,8,8,power_law_1.2,0.19801599502563477
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,8,8,power_law_1.2,0.19600000381469726
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,8,8,power_law_1.01,2.137976379394531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,8,8,power_law_1.2,0.21357440948486328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,8,8,power_law_1.2,0.24775167465209963
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,8,8,power_law_1.2,0.23942655563354492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,8,8,power_law_1.2,0.2565107154846191
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,8,8,power_law_1.2,0.3051187133789063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,8,8,power_law_1.2,0.09650303840637206
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,8,8,power_law_1.2,0.3500428771972656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,8,8,power_law_1.2,0.425849609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,8,8,power_law_1.2,0.5366476821899414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,8,8,power_law_1.01,2.9656524658203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,8,8,power_law_1.2,0.66229248046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,8,8,power_law_1.2,0.22262655258178712
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,8,8,power_law_1.2,0.9467212677001953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,8,8,power_law_1.2,0.049132800102233885
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,8,8,power_law_1.2,1.1588992309570312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,8,8,power_law_1.2,0.06075903892517089
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,8,8,power_law_1.2,0.06952320098876953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,8,8,power_law_1.2,0.09263232231140137
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,8,8,power_law_1.2,0.0973145580291748
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,8,8,power_law_1.2,1.7769894409179687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,8,8,power_law_1.2,0.11241344451904296
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,8,8,power_law_1.2,0.11086720466613768
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,8,8,power_law_1.2,0.13944831848144532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,8,8,power_law_1.2,0.14018303871154786
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,8,8,power_law_1.2,0.14418815612792968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,8,8,power_law_1.2,0.15726079940795898
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,8,8,power_law_1.2,0.16370943069458008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,8,8,power_law_1.2,0.18386816024780273
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,8,8,power_law_1.2,0.17005311965942382
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,8,8,power_law_1.2,0.10422400474548339
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,8,8,power_law_1.2,0.1791244888305664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,8,8,power_law_1.2,0.20104320526123046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,8,8,power_law_1.2,0.28415231704711913
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,8,8,power_law_1.2,0.23924608230590821
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,8,8,power_law_1.2,0.3883212661743164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,8,8,power_law_1.2,2.2192665100097657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,8,8,power_law_1.2,0.4714009475708008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,8,8,power_law_1.2,0.6576025390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,16,4,balanced,0.036849920749664304
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,8,8,power_law_1.2,0.8458841705322266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,16,4,balanced,0.054329600334167484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,16,4,balanced,0.07917695999145508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,16,4,balanced,0.12373503684997558
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,8,8,power_law_1.2,1.1926783752441406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,16,4,balanced,0.2000217628479004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,16,4,balanced,0.28213504791259764
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,8,8,power_law_1.2,1.5700416564941406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,16,4,balanced,0.29510656356811527
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,16,4,balanced,0.2995327949523926
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,16,4,balanced,0.31122047424316407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,16,4,balanced,0.31584768295288085
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,16,4,balanced,0.3280780792236328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,16,4,balanced,0.3529574584960938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,16,4,balanced,0.3604838562011719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,8,8,power_law_1.2,3.384945983886719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,16,4,balanced,0.3791104125976562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,8,8,power_law_1.2,2.306493377685547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,16,4,balanced,0.38572032928466793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,16,4,balanced,0.401605110168457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,16,4,balanced,0.44408832550048827
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,16,4,balanced,0.49836158752441406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,16,4,balanced,0.7857433319091797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,16,4,balanced,0.866644515991211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,16,4,balanced,0.3711193466186523
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,16,4,balanced,1.2484288024902344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,8,8,power_law_1.2,4.438934936523437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,16,4,balanced,0.17979391098022462
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,16,4,balanced,1.4271463012695311
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,16,4,balanced,0.3278412628173828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,16,4,balanced,0.5462911987304688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,16,4,balanced,2.164305877685547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,16,4,balanced,1.0296614074707031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,16,4,balanced,2.72621826171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,16,4,balanced,1.9118809509277344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,16,4,balanced,3.9984307861328126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,16,4,balanced,2.659918212890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,16,4,balanced,2.623956604003906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,8,8,power_law_1.2,3.040558166503906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,16,4,balanced,2.6376397705078127
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,16,4,balanced,5.09755126953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,16,4,balanced,2.669030456542969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,16,4,balanced,2.6255410766601566
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,16,4,balanced,2.1996287536621093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,16,4,balanced,2.211308746337891
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,16,4,balanced,2.6484478759765624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,16,4,balanced,2.2425535583496092
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,16,4,balanced,2.2251327514648436
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,16,4,balanced,2.27525634765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,16,4,balanced,2.3043788146972655
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,16,4,balanced,2.2208895874023438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,16,4,balanced,2.350777587890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,16,4,balanced,2.4649331665039065
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,16,4,balanced,2.5369421386718747
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,16,4,power_law_1.01,0.10880767822265625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,16,4,power_law_1.01,0.10728320121765136
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,16,4,power_law_1.01,0.1437491226196289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,16,4,power_law_1.01,0.18276607513427734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,16,4,balanced,4.820232849121093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,16,4,power_law_1.01,0.2597555160522461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,16,4,power_law_1.01,0.07181183815002441
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,16,4,power_law_1.01,0.2588083267211914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,16,4,power_law_1.01,0.26695167541503906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,16,4,balanced,4.995582580566406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,16,4,power_law_1.01,0.2886732864379883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,16,4,power_law_1.01,0.313832950592041
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,16,4,power_law_1.01,0.3133951950073242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,16,4,power_law_1.01,0.3508070373535156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,16,4,power_law_1.01,0.3527577590942383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,16,4,power_law_1.01,0.3779865646362305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,16,4,power_law_1.01,0.27738624572753906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,16,4,power_law_1.01,0.42301952362060546
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,16,4,balanced,7.424804077148437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,16,4,power_law_1.01,0.6147084808349609
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,16,4,balanced,7.8292431640624995
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,16,4,power_law_1.01,0.3304972839355469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,16,4,power_law_1.01,0.9100633239746093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,16,4,power_law_1.01,1.586639404296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,16,4,power_law_1.01,0.503494415283203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,16,4,power_law_1.01,2.208057556152344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,16,4,power_law_1.01,0.7566118621826172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,16,4,power_law_1.01,0.5069478225708008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,16,4,power_law_1.01,2.8494232177734373
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,16,4,power_law_1.01,0.9685234832763673
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,16,4,power_law_1.01,1.2419379425048827
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,16,4,power_law_1.01,0.9395852661132812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,16,4,balanced,12.648720703124999
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,16,4,power_law_1.01,1.2782489776611328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,16,4,power_law_1.01,1.6225395202636719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,16,4,power_law_1.01,4.247718505859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,16,4,power_law_1.01,2.4560934448242184
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,16,4,power_law_1.01,2.2409075927734374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,16,4,power_law_1.01,2.3447219848632814
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,16,4,power_law_1.01,5.4410009765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,16,4,balanced,15.46404541015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,16,4,power_law_1.01,2.410387268066406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,16,4,power_law_1.01,2.437533416748047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,16,4,power_law_1.01,2.3410047912597656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,16,4,power_law_1.01,1.941808624267578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,16,4,power_law_1.01,1.993955841064453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,16,4,power_law_1.01,2.0365708923339847
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,16,4,power_law_1.01,2.05957763671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,16,4,power_law_1.01,2.0673983764648436
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,16,4,power_law_1.01,2.1995660400390626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,16,4,power_law_1.01,2.416293182373047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,16,4,power_law_1.01,2.6550695800781248
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,16,4,power_law_1.01,3.6054873657226567
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,16,4,power_law_1.2,0.07176703929901122
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,16,4,power_law_1.2,0.09092991828918458
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,16,4,power_law_1.2,0.10849408149719238
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,16,4,power_law_1.2,0.14737024307250976
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,16,4,power_law_1.2,0.168536319732666
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,16,4,power_law_1.2,0.2393190383911133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,16,4,power_law_1.01,5.582667236328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,16,4,power_law_1.2,0.25220991134643556
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,16,4,power_law_1.2,0.2700019264221191
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,16,4,power_law_1.2,0.27025407791137696
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,16,4,power_law_1.2,0.29397632598876955
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,16,4,power_law_1.2,0.3197542381286621
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,16,4,power_law_1.2,0.3223603057861328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,16,4,power_law_1.2,0.32698879241943357
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,16,4,power_law_1.2,0.37299968719482424
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,16,4,power_law_1.2,0.3799065780639649
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,16,4,power_law_1.01,7.774268188476563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,16,4,power_law_1.2,0.40024063110351565
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,16,4,power_law_1.2,0.44408702850341797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,16,4,power_law_1.2,0.5448038482666016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,16,4,power_law_1.2,0.6305523300170899
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,16,4,power_law_1.2,0.7703488159179688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,16,4,power_law_1.2,0.9343513488769531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,16,4,power_law_1.2,1.3203929138183592
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,16,4,power_law_1.01,9.604751586914062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,16,4,power_law_1.2,1.634324493408203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,16,4,power_law_1.2,2.2718885803222655
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,16,4,power_law_1.2,0.5167193603515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,16,4,power_law_1.2,2.9202099609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,16,4,power_law_1.2,0.6909273529052735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,16,4,power_law_1.01,13.531217041015626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,16,4,power_law_1.2,0.9096844482421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,16,4,power_law_1.2,1.2023078155517577
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,16,4,power_law_1.2,4.301209716796874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,16,4,power_law_1.2,1.512390441894531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,16,4,power_law_1.2,5.546113281249999
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,16,4,power_law_1.2,2.343604431152344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,16,4,power_law_1.2,2.169043273925781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,16,4,power_law_1.2,2.3049754333496093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,16,4,power_law_1.2,2.3690393066406252
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,16,4,power_law_1.01,17.851767578125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,16,4,power_law_1.2,2.302210540771484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,16,4,power_law_1.2,2.3629222106933594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,16,4,power_law_1.2,2.0146304321289064
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,16,4,power_law_1.01,3.1229937744140623
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,16,4,power_law_1.2,2.1577792358398438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,16,4,power_law_1.2,2.190955505371094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,16,4,power_law_1.2,2.222013397216797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,16,4,power_law_1.01,4.612510681152344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,16,4,power_law_1.2,2.3486976623535156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,16,4,power_law_1.2,2.569647216796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,16,4,power_law_1.2,2.7765209960937502
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,16,4,power_law_1.2,3.374403991699219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,16,4,power_law_1.2,3.905592346191406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,32,2,balanced,0.045203199386596685
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,32,2,balanced,0.062415361404418945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,32,2,balanced,0.09298303604125976
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,32,2,balanced,0.15658368110656737
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,16,4,power_law_1.2,4.889985351562499
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,32,2,balanced,0.27376384735107423
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,32,2,balanced,0.3786111831665039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,32,2,balanced,0.3901119995117187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,16,4,power_law_1.2,5.962703247070313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,32,2,balanced,0.3948108673095703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,32,2,balanced,0.4101529693603515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,32,2,balanced,0.419873275756836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,32,2,balanced,0.43524223327636713
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,32,2,balanced,0.409167366027832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,32,2,balanced,0.44462337493896487
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,32,2,balanced,0.4291712188720703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,16,4,power_law_1.2,8.40202392578125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,32,2,balanced,0.47149311065673827
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,32,2,balanced,0.4565657424926758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,32,2,balanced,0.49050750732421877
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,32,2,balanced,0.5258995056152344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,32,2,balanced,0.4388915252685547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,32,2,balanced,0.8506829071044922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,32,2,balanced,0.9244659423828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,16,4,power_law_1.2,10.091471557617187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,32,2,balanced,1.3630694580078124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,32,2,balanced,1.5111474609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,32,2,balanced,2.3414834594726566
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,32,2,balanced,2.9109722900390627
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,16,4,power_law_1.2,14.006959228515626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,32,2,balanced,4.324898681640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,16,4,power_law_1.2,1.9909478759765626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,32,2,balanced,1.1526592254638672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,32,2,balanced,5.411044921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,32,2,balanced,2.1532864379882812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,32,2,balanced,0.20663551330566404
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,32,2,balanced,0.33116798400878905
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,32,2,balanced,3.085857238769531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,32,2,balanced,0.6038963317871093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,32,2,balanced,3.0826483154296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,32,2,balanced,3.1124954223632812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,32,2,balanced,2.9890124511718748
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,32,2,balanced,3.0862579345703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,32,2,balanced,3.1312026977539062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,32,2,balanced,3.086134948730469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,32,2,balanced,3.135278015136719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,32,2,balanced,3.1323175048828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,32,2,balanced,3.0425497436523434
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,32,2,balanced,3.008229064941406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,32,2,balanced,3.0251147460937498
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,32,2,balanced,3.1359576416015624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,32,2,balanced,3.143607177734375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,16,4,power_law_1.2,18.78598876953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,32,2,balanced,3.3018673706054686
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,32,2,balanced,3.3900378417968753
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,32,2,power_law_1.01,0.0606553602218628
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,32,2,power_law_1.01,0.10766336441040039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,32,2,balanced,6.3391796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,32,2,power_law_1.01,0.208538875579834
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,32,2,balanced,6.550763549804688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,32,2,power_law_1.01,0.08882431983947754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,32,2,power_law_1.01,0.3077337646484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,32,2,power_law_1.01,0.33664382934570314
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,32,2,power_law_1.01,0.3711641693115234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,32,2,power_law_1.01,0.33865345001220704
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,32,2,power_law_1.01,0.15881343841552736
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,32,2,balanced,9.582374267578125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,32,2,power_law_1.01,0.4120947265625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,32,2,power_law_1.01,0.3616947174072266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,32,2,power_law_1.01,0.4551129531860352
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,32,2,power_law_1.01,0.4972812652587891
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,32,2,power_law_1.01,0.42074497222900387
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,32,2,power_law_1.01,0.566250228881836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,32,2,power_law_1.01,0.47012863159179685
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,32,2,power_law_1.01,0.5391564941406249
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,32,2,power_law_1.01,0.6563648223876953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,32,2,balanced,10.339373779296874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,32,2,power_law_1.01,0.7959552001953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,32,2,power_law_1.01,0.43851776123046876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,32,2,power_law_1.01,0.9839833831787109
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,32,2,power_law_1.01,1.3230911254882813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,32,2,power_law_1.01,1.669354248046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,32,2,power_law_1.01,0.32512256622314456
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,32,2,power_law_1.01,0.5936883163452149
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,32,2,power_law_1.01,0.7857408142089843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,32,2,power_law_1.01,2.988849792480469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,32,2,power_law_1.01,2.303023376464844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,32,2,power_law_1.01,1.5420748901367187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,32,2,power_law_1.01,4.4008984374999995
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,32,2,power_law_1.01,2.4238259887695315
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,32,2,power_law_1.01,5.696015624999999
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,32,2,power_law_1.01,2.5145228576660155
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,32,2,balanced,16.500806884765627
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,32,2,power_law_1.01,2.645646057128906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,32,2,power_law_1.01,2.6551858520507814
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,32,2,power_law_1.01,2.7189364624023438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,32,2,power_law_1.01,1.1388851165771485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,32,2,power_law_1.01,2.7548248291015627
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,32,2,power_law_1.01,2.8323443603515623
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,32,2,power_law_1.01,2.8532070922851562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,32,2,power_law_1.01,2.9194430541992187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,32,2,power_law_1.01,2.8175143432617187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,32,2,balanced,20.5327685546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,32,2,power_law_1.01,2.7964453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,32,2,power_law_1.01,2.863507080078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,32,2,power_law_1.01,3.1425421142578123
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,32,2,power_law_1.01,3.4296948242187497
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,32,2,power_law_1.01,4.044885864257813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,32,2,power_law_1.01,4.61607666015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,32,2,power_law_1.2,0.060660481452941895
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,32,2,power_law_1.2,0.08001664161682129
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,32,2,power_law_1.2,0.10792575836181642
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,32,2,power_law_1.01,5.800806274414063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,32,2,power_law_1.2,0.1890764808654785
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,32,2,power_law_1.2,0.32221183776855467
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,32,2,power_law_1.2,0.31324928283691406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,32,2,power_law_1.2,0.33299713134765624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,32,2,power_law_1.2,0.3640447998046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,32,2,power_law_1.01,7.146176147460937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,32,2,power_law_1.2,0.38423038482666017
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,32,2,power_law_1.2,0.1475379180908203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,32,2,power_law_1.2,0.4095756912231446
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,32,2,power_law_1.2,0.46928897857666013
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,32,2,power_law_1.2,0.5035878372192383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,32,2,power_law_1.2,0.4316185760498047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,32,2,power_law_1.2,0.5770124816894532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,32,2,power_law_1.2,0.4496345520019531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,32,2,power_law_1.2,0.48183425903320315
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,32,2,power_law_1.2,0.5465574264526367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,32,2,power_law_1.2,0.6677311706542969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,32,2,power_law_1.2,0.8109056091308593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,32,2,power_law_1.2,1.0024742126464843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,32,2,power_law_1.01,9.380228881835937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,32,2,power_law_1.2,1.3545741271972656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,32,2,power_law_1.2,1.7016050720214841
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,32,2,power_law_1.2,0.326324462890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,32,2,power_law_1.2,2.360801239013672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,32,2,power_law_1.2,0.49863937377929685
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,32,2,power_law_1.01,11.77565185546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,32,2,power_law_1.2,3.0383309936523437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,32,2,power_law_1.2,0.778703384399414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,32,2,power_law_1.2,1.0583142089843751
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,32,2,power_law_1.2,4.4660186767578125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,32,2,power_law_1.2,1.4441958618164061
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,32,2,power_law_1.2,2.4166015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,32,2,power_law_1.2,2.6548672485351563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,32,2,power_law_1.2,5.722173461914062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,32,2,power_law_1.2,2.6000704956054688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,32,2,power_law_1.2,2.637469482421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,32,2,power_law_1.2,2.6972698974609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,32,2,power_law_1.2,2.8199667358398437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,32,2,power_law_1.01,16.803321533203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,32,2,power_law_1.2,2.8782131958007815
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,32,2,power_law_1.2,2.8911526489257815
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,32,2,power_law_1.2,3.08606201171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,32,2,power_law_1.2,2.8520895385742184
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,32,2,power_law_1.2,2.91383544921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,32,2,power_law_1.2,3.0881088256835936
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,32,2,power_law_1.2,3.372362365722656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,32,2,power_law_1.2,3.6389120483398436
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,32,2,power_law_1.2,4.337401733398438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,32,2,power_law_1.2,4.815897521972657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,16,8,balanced,0.023724799156188962
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,16,8,balanced,0.023733758926391603
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,16,8,balanced,0.023676159381866454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,16,8,balanced,0.03674623966217041
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,16,8,balanced,0.037241599559783935
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,16,8,balanced,0.0374015998840332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,16,8,balanced,0.03795072078704834
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,32,2,power_law_1.01,21.879052734375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,16,8,balanced,0.03898112058639526
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,16,8,balanced,0.03917567968368531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,16,8,balanced,0.039290881156921385
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,16,8,balanced,0.04008959770202637
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,32,2,power_law_1.2,5.92385009765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,16,8,balanced,0.04042367935180664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,16,8,balanced,0.04171775817871094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,16,8,balanced,0.04477695941925049
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,16,8,balanced,0.04514175891876221
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,16,8,balanced,0.04670976161956787
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,16,8,balanced,0.03841919898986816
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,16,8,balanced,0.05528575897216796
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,16,8,balanced,0.06279551982879639
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,16,8,balanced,0.08391551971435547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,16,8,balanced,0.08055168151855469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,16,8,balanced,0.0981158447265625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,16,8,balanced,0.11150976181030274
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,16,8,balanced,0.17372928619384767
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,16,8,balanced,0.22054912567138674
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,16,8,balanced,0.3238131332397461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,16,8,balanced,0.024573440551757815
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,16,8,balanced,0.39988094329833984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,16,8,balanced,0.02439807891845703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,32,2,power_law_1.2,7.35064697265625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,16,8,balanced,0.024675838947296143
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,16,8,balanced,0.041070079803466795
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,16,8,balanced,0.04116608142852783
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,16,8,balanced,0.04154240131378174
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,16,8,balanced,0.0416153621673584
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,16,8,balanced,0.041821441650390624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,16,8,balanced,0.042133760452270505
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,16,8,balanced,0.04225279808044434
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,16,8,balanced,0.0424396800994873
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,16,8,balanced,0.04285696029663086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,16,8,balanced,0.04345727920532226
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,16,8,balanced,0.043842558860778806
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,16,8,balanced,0.044556798934936526
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,16,8,balanced,0.04592639923095703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,16,8,balanced,0.047779841423034666
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,16,8,balanced,0.049616641998291015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,16,8,balanced,0.059528961181640624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,16,8,balanced,0.053858561515808104
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,16,8,balanced,0.061015038490295416
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,16,8,balanced,0.04143231868743896
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,16,8,balanced,0.08509311676025391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,16,8,balanced,0.10452992439270019
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,16,8,balanced,0.15774720191955566
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,16,8,balanced,0.18650880813598633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,16,8,power_law_1.01,0.03577343940734863
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,16,8,power_law_1.01,0.03680255889892578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,16,8,power_law_1.01,0.03692800045013428
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,16,8,balanced,0.07186560153961181
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,16,8,power_law_1.01,0.037628159523010254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,16,8,power_law_1.01,0.03907327890396119
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,16,8,power_law_1.01,0.039650559425354004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,16,8,power_law_1.01,0.0408678388595581
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,16,8,power_law_1.01,0.02363264083862305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,16,8,power_law_1.01,0.042643198966979975
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,16,8,power_law_1.01,0.04808576107025146
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,32,2,power_law_1.2,9.796845092773438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,16,8,power_law_1.01,0.050460162162780764
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,16,8,power_law_1.01,0.056572160720825194
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,16,8,power_law_1.01,0.06425727844238281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,16,8,power_law_1.01,0.05893887996673584
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,16,8,power_law_1.01,0.0775871992111206
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,16,8,power_law_1.01,0.08397184371948242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,16,8,power_law_1.01,0.08306431770324707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,16,8,power_law_1.01,0.09415295600891113
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,16,8,power_law_1.01,0.10300800323486328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,16,8,power_law_1.01,0.14537856101989746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,16,8,power_law_1.01,0.25074047088623047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,16,8,power_law_1.01,0.21444480895996093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,16,8,power_law_1.01,0.8298611450195313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,16,8,power_law_1.01,0.02463871955871582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,32,2,power_law_1.2,12.39955322265625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,16,8,power_law_1.01,0.040767998695373536
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,16,8,power_law_1.01,0.041080322265625004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,16,8,power_law_1.01,0.041483521461486816
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,16,8,power_law_1.01,0.04152959823608399
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,16,8,power_law_1.01,0.04250624179840088
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,16,8,power_law_1.01,0.04321023941040039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,16,8,power_law_1.01,0.04361728191375732
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,16,8,power_law_1.01,0.04409088134765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,16,8,power_law_1.01,0.04520576000213623
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,16,8,power_law_1.01,0.04224639892578125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,16,8,power_law_1.01,0.04635136127471924
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,16,8,power_law_1.01,0.04742656230926513
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,16,8,power_law_1.01,0.05294335842132568
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,16,8,power_law_1.01,0.04919424057006836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,16,8,power_law_1.01,0.05847296237945556
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,16,8,power_law_1.01,0.05687935829162598
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,16,8,power_law_1.01,0.059749121665954585
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,16,8,power_law_1.01,0.38329601287841797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,16,8,power_law_1.01,0.07122432231903077
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,16,8,power_law_1.01,0.08051584243774415
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,16,8,power_law_1.01,0.5388159942626953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,16,8,power_law_1.01,0.6957107543945312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,32,2,power_law_1.2,17.498577880859376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,16,8,power_law_1.01,0.10090623855590822
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,16,8,power_law_1.01,0.15316224098205566
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,16,8,power_law_1.2,0.023919360637664793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,16,8,power_law_1.2,0.03569407939910889
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,16,8,power_law_1.2,0.03671808004379272
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,16,8,power_law_1.2,0.03686784029006958
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,16,8,power_law_1.2,0.03757312059402466
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,16,8,power_law_1.2,0.03878911972045899
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,16,8,power_law_1.2,0.039997439384460445
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,16,8,power_law_1.2,0.04102911949157715
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,16,8,power_law_1.01,0.18428031921386717
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,16,8,power_law_1.2,0.04306816101074219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,16,8,power_law_1.2,0.05268991947174072
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,16,8,power_law_1.2,0.05624063968658447
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,16,8,power_law_1.2,0.05837696075439454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,16,8,power_law_1.2,0.04806655883789063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,16,8,power_law_1.2,0.06535679817199706
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,16,8,power_law_1.2,0.07817471981048583
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,16,8,power_law_1.2,0.08187775611877442
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,16,8,power_law_1.01,0.3274483108520508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,16,8,power_law_1.2,0.08037247657775878
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,16,8,power_law_1.01,0.25949951171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,16,8,power_law_1.2,0.09298944473266602
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,16,8,power_law_1.2,0.1457817554473877
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,16,8,power_law_1.2,0.11168640136718751
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,16,8,power_law_1.2,0.2127257537841797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,16,8,power_law_1.01,1.3107455444335936
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,16,8,power_law_1.2,0.29443328857421874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,16,8,power_law_1.01,0.5325030517578125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,16,8,power_law_1.2,0.024559359550476074
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,16,8,power_law_1.2,0.35877246856689454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,16,8,power_law_1.2,0.04066688060760498
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,16,8,power_law_1.2,0.04107647895812988
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,32,2,power_law_1.2,22.95583251953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,16,8,power_law_1.2,0.04114816188812256
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,16,8,power_law_1.2,0.041350398063659664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,16,8,power_law_1.2,0.042442240715026856
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,16,8,power_law_1.2,0.043128318786621093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,16,8,power_law_1.2,0.04360832214355469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,16,8,power_law_1.2,0.04442624092102051
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,16,8,power_law_1.2,0.04566271781921387
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,16,8,power_law_1.2,0.04654719829559326
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,16,8,power_law_1.2,0.04210559844970703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,16,8,power_law_1.2,0.04745471954345703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,16,8,power_law_1.2,0.04954239845275879
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,16,8,power_law_1.2,0.517580795288086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,16,8,power_law_1.2,0.053958401679992676
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,16,8,power_law_1.2,0.05900928020477295
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,16,8,power_law_1.2,0.060211200714111325
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,16,8,power_law_1.2,0.05583231925964356
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,16,8,power_law_1.2,0.7232934570312499
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,16,8,power_law_1.2,0.07185023784637451
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,16,8,power_law_1.2,0.08344191551208496
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,16,8,power_law_1.2,0.10051072120666504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,16,8,power_law_1.2,0.17384960174560546
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,16,8,power_law_1.2,0.14573056221008301
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,32,4,balanced,0.022196478843688965
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,32,4,balanced,0.022159359455108642
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,32,4,balanced,0.02380160093307495
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,32,4,balanced,0.0340339207649231
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,32,4,balanced,0.03479167938232422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,32,4,balanced,0.03513344049453736
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,16,8,power_law_1.2,0.22197504043579103
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,32,4,balanced,0.03569664001464844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,32,4,balanced,0.03607295989990235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,32,4,balanced,0.0372710394859314
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,32,4,balanced,0.037634561061859126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,32,4,balanced,0.03868160009384155
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,32,4,balanced,0.03929728031158447
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,32,4,balanced,0.039870719909667965
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,32,4,balanced,0.043297281265258794
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,32,4,balanced,0.045780482292175295
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,32,4,balanced,0.046196479797363285
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,32,4,balanced,0.05589888095855713
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,32,4,balanced,0.07806719779968262
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,32,4,balanced,0.07428095817565919
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,32,4,balanced,0.0844172763824463
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,32,4,balanced,0.09369215965270997
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,32,4,balanced,0.11268608093261719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,32,4,balanced,0.1296281623840332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,32,4,balanced,0.195546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,32,4,balanced,0.25142143249511717
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,32,4,balanced,0.37478912353515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,32,4,balanced,0.023404800891876222
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,32,4,balanced,0.4701824188232422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,16,8,power_law_1.2,1.0276083374023437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,32,4,balanced,0.039882240295410154
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,32,4,balanced,0.03982464075088501
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,32,4,balanced,0.04021376132965088
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,32,4,balanced,0.04070528030395508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,32,4,balanced,0.04049536228179932
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,16,8,power_law_1.01,0.6363353729248047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,32,4,balanced,0.0408128023147583
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,32,4,balanced,0.023403520584106448
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,32,4,balanced,0.04148096084594727
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,32,4,balanced,0.04164224147796631
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,32,4,balanced,0.04259071826934814
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,32,4,balanced,0.04337535858154297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,32,4,balanced,0.04437376022338867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,32,4,balanced,0.04539648056030273
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,32,4,balanced,0.04841216087341309
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,32,4,balanced,0.05666944026947022
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,32,4,balanced,0.05232511997222901
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,32,4,balanced,0.058311681747436526
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,32,4,balanced,0.06834943771362305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,32,4,balanced,0.078821120262146
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,32,4,balanced,0.09482111930847167
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,32,4,balanced,0.04059135913848877
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,32,4,balanced,0.12627200126647947
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,32,4,balanced,0.15435263633728027
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,32,4,balanced,0.23122432708740231
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,32,4,power_law_1.01,0.022228479385375977
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,32,4,balanced,0.28981632232666016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,16,8,power_law_1.2,1.425945587158203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,32,4,power_law_1.01,0.023987200260162354
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,32,4,power_law_1.01,0.03429888010025024
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,32,4,power_law_1.01,0.03442431926727295
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,32,4,power_law_1.01,0.03496063947677612
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,32,4,power_law_1.01,0.03760256052017212
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,32,4,power_law_1.01,0.03927551984786987
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,32,4,power_law_1.01,0.04136320114135742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,32,4,power_law_1.01,0.04042496204376221
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,32,4,power_law_1.01,0.04248703956604004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,32,4,power_law_1.01,0.046558718681335445
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,32,4,power_law_1.01,0.04651264190673828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,32,4,power_law_1.01,0.05816448211669922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,32,4,power_law_1.01,0.07326848030090331
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,32,4,balanced,0.025125119686126705
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,32,4,power_law_1.01,0.07519231796264648
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,32,4,power_law_1.01,0.07947904109954834
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,32,4,power_law_1.01,0.0863424015045166
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,32,4,power_law_1.01,0.10066944122314454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,32,4,power_law_1.01,0.11306879997253419
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,32,4,power_law_1.01,0.03568511962890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,32,4,power_law_1.01,0.1393062400817871
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,32,4,power_law_1.01,0.20901248931884764
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,32,4,power_law_1.01,0.6286553573608399
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,32,4,power_law_1.01,0.023400959968566896
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,32,4,power_law_1.01,0.2721164894104004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,32,4,power_law_1.01,0.024706559181213377
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,32,4,power_law_1.01,0.03959424018859863
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,32,4,power_law_1.01,0.34773502349853513
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,32,4,power_law_1.01,0.03967871904373169
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,32,4,power_law_1.01,0.03983360052108765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,32,4,power_law_1.01,0.040317440032958986
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,32,4,power_law_1.01,0.040816640853881835
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,32,4,power_law_1.01,0.04157311916351318
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,32,4,power_law_1.01,0.04206079959869385
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,32,4,power_law_1.01,0.04245120048522949
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,32,4,power_law_1.01,0.043884801864624026
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,32,4,power_law_1.01,0.044801278114318846
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,32,4,power_law_1.01,0.49464447021484376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,32,4,power_law_1.01,0.04582655906677246
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,32,4,power_law_1.01,0.054112000465393065
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,32,4,power_law_1.01,0.047870721817016605
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,32,4,power_law_1.01,0.05397503852844239
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,32,4,power_law_1.01,0.05712895870208741
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,32,4,power_law_1.01,0.057955842018127444
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,16,8,power_law_1.2,0.3613695907592773
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,32,4,power_law_1.01,0.06876031875610351
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,32,4,power_law_1.01,0.08190336227416992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,32,4,power_law_1.01,0.09288064002990723
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,32,4,power_law_1.01,0.1241983985900879
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,32,4,power_law_1.01,0.1661222457885742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,32,4,power_law_1.2,0.022277119159698485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,32,4,power_law_1.2,0.02376192092895508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,32,4,power_law_1.2,0.03402112007141113
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,32,4,power_law_1.2,0.034429440498352046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,16,8,power_law_1.2,0.503614730834961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,32,4,power_law_1.2,0.03495167970657349
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,32,4,power_law_1.2,0.03614975929260254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,32,4,power_law_1.2,0.03751039981842041
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,32,4,power_law_1.2,0.038928639888763425
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,32,4,power_law_1.2,0.03999488115310669
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,32,4,power_law_1.2,0.042639360427856446
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,32,4,power_law_1.2,0.04686463832855224
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,32,4,power_law_1.2,0.05026175975799561
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,32,4,power_law_1.2,0.05147391796112061
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,32,4,power_law_1.2,0.056179199218749996
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,32,4,power_law_1.2,0.07218688011169434
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,32,4,power_law_1.2,0.0774182415008545
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,32,4,power_law_1.2,0.0781388807296753
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,32,4,power_law_1.2,0.08723456382751464
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,32,4,power_law_1.2,0.09590271949768067
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,32,4,power_law_1.2,0.11373696327209473
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,32,4,power_law_1.01,0.21728384017944338
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,32,4,power_law_1.2,0.13975168228149415
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,32,4,power_law_1.01,0.9097561645507813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,32,4,power_law_1.2,0.2060044860839844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,32,4,power_law_1.01,0.3124415969848633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,32,4,power_law_1.01,0.37978496551513674
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,32,4,power_law_1.2,0.02330496072769165
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,16,8,power_law_1.2,0.6298355102539063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,32,4,power_law_1.2,0.43438465118408204
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,32,4,power_law_1.2,0.03957887887954712
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,32,4,power_law_1.2,0.2720256042480469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,32,4,power_law_1.2,0.03978240013122559
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,32,4,power_law_1.2,0.04047359943389893
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,32,4,power_law_1.2,0.040906238555908206
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,32,4,power_law_1.2,0.04159872055053711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,32,4,power_law_1.2,0.04227839946746827
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,32,4,power_law_1.2,0.04290175914764404
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,32,4,power_law_1.2,0.04400000095367432
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,32,4,power_law_1.2,0.04502655982971192
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,32,4,power_law_1.2,0.0458355188369751
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,32,4,power_law_1.2,0.048080639839172365
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,32,4,power_law_1.2,0.025098240375518797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,32,4,power_law_1.2,0.050945281982421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,32,4,power_law_1.2,0.05583231925964356
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,32,4,power_law_1.2,0.05818751811981201
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,32,4,power_law_1.2,0.05466112136840821
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,32,4,power_law_1.2,0.3532134246826172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,32,4,power_law_1.2,0.06850687980651855
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,32,4,power_law_1.2,0.08939392089843749
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,32,4,power_law_1.2,0.0817728042602539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,32,4,power_law_1.2,0.03943167924880982
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,32,4,power_law_1.2,0.12904447555541992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,32,4,power_law_1.01,0.5356595230102539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,16,8,balanced,0.04379648208618164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,16,8,balanced,0.043878397941589355
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,16,8,balanced,0.029671680927276612
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,16,8,balanced,0.05242496013641358
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,16,8,balanced,0.050392317771911624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,16,8,balanced,0.04865536212921143
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,16,8,balanced,0.0516428804397583
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,16,8,balanced,0.05881599903106689
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,16,8,balanced,0.07461247920989991
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,32,4,power_law_1.2,0.16073728561401368
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,16,8,balanced,0.08079744338989257
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,16,8,balanced,0.08751872062683105
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,16,8,balanced,0.08852736473083496
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,16,8,balanced,0.08848640441894531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,16,8,balanced,0.09227775573730469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,16,8,balanced,0.09410304069519043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,16,8,balanced,0.07729919910430907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,16,8,balanced,0.09858688354492187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,16,8,balanced,0.10659711837768555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,16,8,balanced,0.10998016357421876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,16,8,balanced,0.12686335563659668
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,16,8,balanced,0.12721920013427734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,16,8,balanced,0.14963199615478515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,16,8,balanced,0.16979455947875977
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,16,8,balanced,0.2848512077331543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,16,8,balanced,0.33491455078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,16,8,balanced,0.4831398391723633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,32,4,power_law_1.2,0.22033279418945312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,16,8,balanced,0.028901119232177735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,16,8,balanced,0.6464256286621094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,16,8,balanced,0.028949759006500243
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,16,8,balanced,0.028805119991302492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,16,8,balanced,0.051038718223571776
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,16,8,balanced,0.050708479881286614
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,16,8,balanced,0.05114751815795898
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,16,8,balanced,0.051545600891113284
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,16,8,balanced,0.051325440406799316
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,16,8,balanced,0.05167359828948974
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,16,8,balanced,0.05196671962738038
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,16,8,balanced,0.05170688152313232
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,16,8,balanced,0.05225344181060791
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,16,8,balanced,0.0528550386428833
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,16,8,balanced,0.05309951782226563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,16,8,balanced,0.05378943920135498
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,16,8,balanced,0.0546342420578003
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,16,8,balanced,0.05611135959625244
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,16,8,balanced,0.060945920944213874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,16,8,balanced,0.06707583904266358
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,16,8,balanced,0.07417727947235107
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,16,8,balanced,0.06939775943756103
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,16,8,balanced,0.08436223983764649
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,16,8,balanced,0.09241984367370606
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,16,8,balanced,0.11816703796386718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,16,8,balanced,0.1459763240814209
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,16,8,balanced,0.2187545585632324
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,16,8,power_law_1.01,0.04570623874664307
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,16,8,power_law_1.01,0.05147520065307617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,16,8,power_law_1.01,0.0809267234802246
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,16,8,power_law_1.01,0.050402560234069825
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,16,8,power_law_1.01,0.08255359649658203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,16,8,power_law_1.01,0.0575654411315918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,16,8,power_law_1.01,0.08944000244140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,16,8,power_law_1.01,0.07333119869232177
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,16,8,power_law_1.01,0.09482111930847167
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,16,8,power_law_1.01,0.09078783988952636
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,16,8,balanced,0.2648204803466797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,16,8,power_law_1.01,0.0997593593597412
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,16,8,power_law_1.01,0.10132736206054686
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,16,8,power_law_1.01,0.10341631889343261
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,16,8,power_law_1.01,0.10790016174316405
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,16,8,power_law_1.01,0.11548543930053712
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,16,8,power_law_1.01,0.12026880264282228
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,16,8,power_law_1.01,0.12744704246520996
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,16,8,power_law_1.01,0.14197888374328613
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,16,8,power_law_1.01,0.1605196762084961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,32,4,power_law_1.2,0.7290214538574219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,16,8,power_law_1.01,0.3175155258178711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,16,8,power_law_1.01,0.26683008193969726
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,32,4,power_law_1.2,0.44594558715820315
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,16,8,power_law_1.01,0.45954433441162107
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,32,4,power_law_1.2,0.3054425621032715
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,16,8,power_law_1.01,0.5970867156982422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,16,8,power_law_1.01,0.028898561000823976
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,16,8,power_law_1.01,0.05033984184265137
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,16,8,power_law_1.01,0.0508454418182373
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,16,8,power_law_1.01,0.0510207986831665
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,16,8,power_law_1.01,0.05113344192504883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,16,8,power_law_1.01,0.051589121818542485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,16,8,power_law_1.01,0.05204095840454102
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,16,8,power_law_1.01,0.05268735885620117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,16,8,power_law_1.01,0.05306367874145508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,16,8,power_law_1.01,0.053569278717041015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,16,8,power_law_1.01,0.05511936187744141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,16,8,power_law_1.01,0.056248321533203124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,16,8,power_law_1.01,0.059169278144836426
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,16,8,power_law_1.01,0.06417791843414307
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,16,8,power_law_1.01,0.06773375988006591
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,16,8,power_law_1.01,0.07047039985656738
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,16,8,power_law_1.01,0.06968063831329345
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,16,8,power_law_1.01,0.08356096267700194
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,16,8,power_law_1.01,0.09041664123535156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,32,4,power_law_1.2,0.9149158477783204
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,16,8,power_law_1.01,0.8324582672119141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,16,8,power_law_1.01,1.012606735229492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,16,8,power_law_1.01,0.11866239547729492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,16,8,power_law_1.01,0.1393113613128662
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,16,8,power_law_1.01,0.18920064926147462
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,16,8,power_law_1.01,0.35149951934814455
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,16,8,power_law_1.01,0.24340351104736327
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,16,8,power_law_1.2,0.04287744045257568
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,16,8,power_law_1.2,0.06448895931243896
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,16,8,power_law_1.2,0.048124160766601566
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,16,8,power_law_1.2,0.055256319046020505
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,16,8,power_law_1.2,0.051454720497131345
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,16,8,power_law_1.2,0.06683519840240479
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,16,8,power_law_1.2,0.07954944133758544
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,16,8,power_law_1.2,0.07863935947418213
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,16,8,power_law_1.2,0.09351679801940918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,16,8,power_law_1.2,0.09363712310791016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,16,8,power_law_1.2,0.09416192054748536
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,16,8,power_law_1.2,0.10136447906494142
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,16,8,power_law_1.01,1.8572979736328126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,16,8,power_law_1.2,0.10416255950927736
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,16,8,power_law_1.2,0.10835455894470214
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,16,8,power_law_1.2,0.115731201171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,16,8,power_law_1.01,0.7073049926757813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,16,8,power_law_1.2,0.12363136291503907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,16,8,power_law_1.2,0.1271065616607666
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,32,4,power_law_1.2,0.548985595703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,16,8,power_law_1.2,0.15047295570373537
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,16,8,power_law_1.2,0.1618547248840332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,16,8,power_law_1.01,0.46969345092773435
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,16,8,power_law_1.2,0.2639948844909668
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,16,8,power_law_1.2,0.3157529640197754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,16,8,power_law_1.01,0.9614399719238282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,16,8,power_law_1.2,0.028840959072113037
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,16,8,power_law_1.2,0.050575361251831055
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,16,8,power_law_1.2,0.05119359970092774
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,16,8,power_law_1.2,0.05122943878173828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,16,8,power_law_1.2,0.05158783912658691
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,16,8,power_law_1.2,0.051870718002319335
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,16,8,power_law_1.2,0.05247360229492187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,16,8,power_law_1.2,0.05324672222137451
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,16,8,power_law_1.2,0.053726720809936526
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,16,8,power_law_1.2,0.05442304134368896
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,16,8,power_law_1.2,0.05616896152496338
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,16,8,power_law_1.2,0.0570636796951294
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,16,8,power_law_1.2,0.05994624137878418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,16,8,power_law_1.2,0.48082942962646485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,16,8,power_law_1.2,0.06388607978820801
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,16,8,power_law_1.2,0.06911359786987305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,16,8,power_law_1.2,0.07154047966003418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,16,8,power_law_1.2,0.07201151847839356
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,16,8,power_law_1.2,0.08408063888549805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,16,8,power_law_1.2,0.09156479835510253
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,16,8,power_law_1.2,0.6148070526123047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,16,8,power_law_1.2,0.9205734252929687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,16,8,power_law_1.2,0.11857279777526855
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,16,8,power_law_1.2,0.14156671524047854
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,16,8,power_law_1.2,1.1698099517822267
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,16,8,power_law_1.2,0.2169036865234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,16,8,power_law_1.2,0.2561548805236816
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,32,4,balanced,0.02809727907180786
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,16,8,power_law_1.2,2.249610290527344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,32,4,balanced,0.0280294394493103
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,32,4,balanced,0.03584383964538575
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,32,4,balanced,0.050032639503479005
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,32,4,balanced,0.0552614402770996
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,32,4,balanced,0.0606873607635498
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,32,4,balanced,0.06305920124053956
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,32,4,balanced,0.06443007946014404
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,32,4,balanced,0.06947711944580079
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,32,4,balanced,0.07089663982391357
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,32,4,balanced,0.059078397750854494
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,32,4,balanced,0.07367680072784424
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,32,4,balanced,0.0805081558227539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,32,4,balanced,0.09116288185119628
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,32,4,balanced,0.09317119598388672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,32,4,balanced,0.09815296173095703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,32,4,balanced,0.10358016014099121
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,32,4,balanced,0.11691519737243652
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,32,4,balanced,0.1130073642730713
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,32,4,balanced,0.12922880172729492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,32,4,balanced,0.13982208251953127
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,32,4,balanced,0.16853120803833008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,32,4,balanced,0.1913382339477539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,32,4,balanced,0.3136191940307617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,32,4,balanced,0.3847603225708008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,16,8,power_law_1.2,0.48181121826171874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,32,4,balanced,0.02732928037643433
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,32,4,balanced,0.7502694702148437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,32,4,balanced,0.02745599985122681
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,32,4,balanced,0.02926464080810547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,32,4,balanced,0.04853888034820557
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,32,4,balanced,0.04884992122650146
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,32,4,balanced,0.04878079891204834
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,16,8,power_law_1.01,2.4489407348632812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,32,4,balanced,0.04927360057830811
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,32,4,balanced,0.051211519241333006
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,32,4,balanced,0.04975103855133057
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,32,4,balanced,0.5673241424560547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,16,8,power_law_1.2,0.3808729553222656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,32,4,balanced,0.0507097578048706
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,32,4,balanced,0.05185279846191406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,32,4,balanced,0.05276544094085693
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,32,4,balanced,0.05937535762786865
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,32,4,balanced,0.06383872032165527
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,32,4,balanced,0.07182975769042968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,32,4,balanced,0.06606592178344727
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,32,4,balanced,0.07772672176361084
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,32,4,balanced,0.08569215774536133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,32,4,balanced,0.049176321029663086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,32,4,balanced,0.10619008064270019
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,32,4,balanced,0.1266495990753174
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,32,4,balanced,0.1694054412841797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,32,4,balanced,0.051175680160522465
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,32,4,balanced,0.20953344345092773
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,16,8,power_law_1.2,1.687230682373047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,32,4,balanced,0.05587456226348877
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,32,4,balanced,0.31916032791137694
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,32,4,power_law_1.01,0.03994368076324463
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,32,4,balanced,0.40137599945068364
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,32,4,power_law_1.01,0.05293439865112305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,32,4,power_law_1.01,0.05530496120452881
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,32,4,power_law_1.01,0.07047679901123047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,32,4,power_law_1.01,0.07665919780731201
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,32,4,power_law_1.01,0.07403264045715333
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,32,4,power_law_1.01,0.08474111557006836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,32,4,power_law_1.01,0.08383999824523926
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,32,4,power_law_1.01,0.09057920455932618
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,32,4,power_law_1.01,0.09242112159729003
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,32,4,power_law_1.01,0.09746432304382324
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,32,4,power_law_1.01,0.10267775535583497
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,32,4,power_law_1.01,0.029446399211883544
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,32,4,power_law_1.01,0.1087168025970459
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,32,4,power_law_1.01,0.060602879524230956
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,32,4,power_law_1.01,0.11328127861022949
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,32,4,power_law_1.01,0.11943167686462401
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,32,4,power_law_1.01,0.14012928009033204
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,32,4,power_law_1.01,0.13021696090698243
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,32,4,power_law_1.01,0.1987046432495117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,32,4,power_law_1.01,0.17932287216186524
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,32,4,power_law_1.01,0.32173694610595704
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,32,4,power_law_1.01,0.02737407922744751
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,32,4,power_law_1.01,0.0290393590927124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,32,4,power_law_1.01,0.04849535942077637
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,32,4,power_law_1.01,0.048719358444213864
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,32,4,power_law_1.01,0.04894591808319092
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,32,4,power_law_1.01,0.0493068790435791
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,32,4,power_law_1.01,0.04985472202301026
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,32,4,power_law_1.01,0.05032063961029053
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,32,4,power_law_1.01,0.050773758888244626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,32,4,power_law_1.01,0.05147520065307617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,32,4,power_law_1.01,0.05289472103118896
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,32,4,power_law_1.01,0.055453438758850095
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,32,4,power_law_1.01,0.058174719810485845
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,32,4,power_law_1.01,0.06147071838378906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,32,4,power_law_1.01,0.37037822723388675
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,32,4,power_law_1.01,0.06656640052795411
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,32,4,power_law_1.01,0.6287884902954102
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,32,4,power_law_1.01,0.06936448097229005
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,32,4,power_law_1.01,0.07153791904449462
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,32,4,power_law_1.01,0.08977536201477052
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,32,4,power_law_1.01,0.8289778900146484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,32,4,power_law_1.01,0.07702911853790283
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,16,8,power_law_1.2,0.9707199859619141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,16,8,power_law_1.2,0.6869670104980469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,32,4,power_law_1.01,0.10989952087402344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,32,4,power_law_1.01,0.12668160438537596
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,32,4,power_law_1.01,0.16898944854736328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,32,4,power_law_1.2,0.03390079975128174
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,32,4,power_law_1.2,0.03586944103240967
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,32,4,power_law_1.2,0.0565555191040039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,32,4,power_law_1.2,0.04627967834472656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,32,4,power_law_1.2,0.047489280700683596
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,32,4,power_law_1.2,0.061687040328979495
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,32,4,power_law_1.2,0.06680319786071778
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,32,4,power_law_1.2,0.07166207790374755
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,32,4,power_law_1.01,0.23274112701416017
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,32,4,power_law_1.2,0.08113151550292969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,32,4,power_law_1.2,0.08272768020629882
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,32,4,power_law_1.2,0.08646528244018555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,32,4,power_law_1.2,0.09308032035827637
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,32,4,power_law_1.2,0.09757696151733398
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,32,4,power_law_1.2,0.10387328147888183
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,32,4,power_law_1.2,0.10940032005310059
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,32,4,power_law_1.2,0.11668224334716797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,32,4,power_law_1.01,1.5607769775390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,32,4,power_law_1.2,0.1179481601715088
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,32,4,power_law_1.01,0.4397158432006836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,32,4,power_law_1.2,0.1408243179321289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,32,4,power_law_1.2,0.12996095657348633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,32,4,power_law_1.01,0.5957580947875977
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,32,4,power_law_1.2,0.18276351928710938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,32,4,power_law_1.2,0.21789695739746096
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,32,4,power_law_1.01,0.3224025726318359
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,32,4,power_law_1.2,0.3132953643798828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,32,4,power_law_1.2,0.027283198833465576
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,32,4,power_law_1.2,0.029122560024261473
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,32,4,power_law_1.01,1.154229736328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,32,4,power_law_1.2,0.048519678115844726
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,32,4,power_law_1.2,0.04893055915832519
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,32,4,power_law_1.2,0.049314560890197756
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,32,4,power_law_1.01,0.8406079864501953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,32,4,power_law_1.2,0.04986495971679687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,32,4,power_law_1.2,0.05049088001251221
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,32,4,power_law_1.2,0.05091072082519531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,32,4,power_law_1.2,0.051895041465759274
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,32,4,power_law_1.2,0.05303936004638672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,32,4,power_law_1.2,0.05540736198425293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,32,4,power_law_1.2,0.05894656181335449
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,32,4,power_law_1.2,0.061976318359375004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,32,4,power_law_1.2,0.06727295875549316
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,32,4,power_law_1.2,0.07051775932312013
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,32,4,power_law_1.2,0.3773811340332031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,32,4,power_law_1.2,0.07179520130157471
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,32,4,power_law_1.2,0.048670721054077146
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,32,4,power_law_1.2,0.0847436809539795
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,32,4,power_law_1.2,0.07949056148529052
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,32,4,power_law_1.2,0.1102412796020508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,32,4,power_law_1.2,0.12860799789428712
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,32,4,power_law_1.2,0.1888960075378418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,32,4,power_law_1.2,0.2349375915527344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,32,4,power_law_1.2,0.7387481689453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,1,128,balanced,0.06743167877197266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,1,128,balanced,0.06744063854217529
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,1,128,balanced,0.06750592231750488
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,32,4,power_law_1.2,0.5909094238281251
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,1,128,balanced,0.06729343891143799
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,1,128,balanced,0.0680128002166748
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,1,128,balanced,0.07798272132873535
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,1,128,balanced,0.1350438404083252
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,1,128,balanced,0.13495424270629883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,1,128,balanced,0.1357043170928955
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,1,128,balanced,0.13567872047424318
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,1,128,balanced,0.13599488258361817
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,1,128,balanced,0.13894399642944336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,1,128,balanced,0.138536958694458
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,1,128,balanced,0.1402726364135742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,1,128,balanced,0.141277437210083
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,1,128,balanced,0.14286080360412598
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,1,128,balanced,0.14417920112609864
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,1,128,balanced,0.14888064384460448
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,1,128,balanced,0.15607040405273437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,1,128,balanced,0.1618560028076172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,1,128,balanced,0.1566220760345459
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,1,128,balanced,0.1754956817626953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,1,128,balanced,0.18154495239257812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,1,128,balanced,0.21672960281372072
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,1,128,balanced,0.2637132835388184
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,1,128,balanced,0.39361793518066407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,1,128,balanced,0.03804415941238403
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,1,128,balanced,0.4541606521606445
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,1,128,balanced,0.0392576003074646
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,1,128,balanced,0.039000320434570315
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,1,128,balanced,0.038255360126495365
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,1,128,balanced,0.0395033597946167
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,1,128,balanced,0.049283838272094725
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,1,128,balanced,0.06482944011688233
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,1,128,balanced,0.06520959854125977
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,1,128,balanced,0.06563199996948242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,1,128,balanced,0.0656166410446167
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,1,128,balanced,0.06570240020751952
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,1,128,balanced,0.06639999866485595
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,1,128,balanced,0.06693376064300537
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,1,128,balanced,0.06744063854217529
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,1,128,balanced,0.06800127983093261
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,1,128,balanced,0.06941952228546142
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,1,128,balanced,0.07089536190032959
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,1,128,balanced,0.07303552150726318
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,1,128,balanced,0.07689727783203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,1,128,balanced,0.08298239707946778
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,1,128,balanced,0.07969279766082764
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,1,128,balanced,0.08805760383605957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,1,128,balanced,0.09680255889892578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,1,128,balanced,0.11492992401123048
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,1,128,balanced,0.1300876808166504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,1,128,balanced,0.16355199813842775
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,1,128,balanced,0.20727039337158204
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,1,128,power_law_1.01,0.07882495880126952
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,32,4,power_law_1.2,0.426611213684082
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,1,128,power_law_1.01,0.07812863826751709
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,1,128,power_law_1.01,0.07817344188690185
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,1,128,power_law_1.01,0.0780787181854248
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,1,128,power_law_1.01,0.1348134422302246
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,1,128,power_law_1.01,0.1373196792602539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,1,128,power_law_1.01,0.13851136207580567
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,1,128,power_law_1.01,0.13882368087768554
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,1,128,power_law_1.01,0.1409318447113037
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,1,128,power_law_1.01,0.14242303848266602
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,1,128,power_law_1.01,0.14382847785949707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,1,128,power_law_1.01,0.1459878444671631
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,1,128,power_law_1.01,0.15025279998779298
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,1,128,power_law_1.01,0.15479424476623535
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,1,128,power_law_1.01,0.159934720993042
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,1,128,power_law_1.01,0.16381824493408204
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,1,128,power_law_1.01,0.17052799224853515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,1,128,power_law_1.01,0.17642623901367188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,32,4,power_law_1.2,0.31794559478759765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,1,128,power_law_1.01,0.21763200759887696
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,1,128,power_law_1.01,0.3097318458557129
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,1,128,power_law_1.01,0.3749555206298828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,1,128,power_law_1.01,0.5197273635864258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,32,4,power_law_1.2,1.160732192993164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,1,128,power_law_1.01,0.6874867248535156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,1,128,power_law_1.01,0.04856959819793701
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,1,128,power_law_1.01,0.8999372863769531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,1,128,power_law_1.01,0.049063677787780764
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,1,128,power_law_1.01,0.04854144096374512
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,1,128,power_law_1.01,1.2981913757324217
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,32,4,power_law_1.2,1.5201805114746094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,1,128,power_law_1.01,0.04818816184997558
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,1,128,power_law_1.01,0.06485119819641114
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,1,128,power_law_1.01,0.06589951992034912
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,1,128,power_law_1.01,0.06856319904327393
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,1,128,power_law_1.01,0.06635263919830323
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,1,128,power_law_1.01,0.06719359874725342
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,1,128,power_law_1.01,0.06831744194030762
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,1,128,power_law_1.01,0.07222655773162842
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,1,128,power_law_1.01,0.0702348804473877
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,1,128,power_law_1.01,0.07635456085205078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,1,128,power_law_1.01,0.08066047668457031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,1,128,power_law_1.01,0.08370559692382813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,1,128,power_law_1.01,0.08636544227600099
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,1,128,power_law_1.01,2.2397196960449217
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,1,128,power_law_1.01,0.09034624099731445
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,1,128,power_law_1.01,0.09863807678222655
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,1,128,power_law_1.01,0.13414015769958496
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,1,128,power_law_1.01,0.14403200149536133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,1,128,power_law_1.01,0.23068544387817386
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,1,128,power_law_1.01,0.07420544147491455
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,1,128,power_law_1.01,0.2737830352783203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,1,128,power_law_1.01,2.0672064208984375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,1,128,power_law_1.01,0.46684543609619145
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,1,128,power_law_1.2,0.07883776187896728
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,1,128,power_law_1.01,0.534961929321289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,1,128,power_law_1.2,0.0780454397201538
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,1,128,power_law_1.2,0.07842432022094727
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,1,128,power_law_1.2,0.129879035949707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,1,128,power_law_1.2,0.1351116752624512
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,1,128,power_law_1.2,0.13770496368408203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,32,4,power_law_1.2,0.6085286331176758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,1,128,power_law_1.2,0.13772159576416015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,1,128,power_law_1.2,0.14011648178100586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,1,128,power_law_1.2,0.1431603240966797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,1,128,power_law_1.2,0.14148608207702637
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,1,128,power_law_1.2,0.14546560287475585
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,1,128,power_law_1.2,0.14413951873779296
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,1,128,power_law_1.2,0.15516160011291505
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,1,128,power_law_1.2,0.15675392150878906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,1,128,power_law_1.2,0.1652249526977539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,1,128,power_law_1.2,0.16926975250244142
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,1,128,power_law_1.2,0.1761612892150879
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,1,128,power_law_1.2,0.2086476707458496
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,1,128,power_law_1.01,0.8827776336669922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,1,128,power_law_1.2,0.26139648437499996
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,1,128,power_law_1.2,0.3704908752441406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,1,128,power_law_1.01,1.0277299499511718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,1,128,power_law_1.2,0.543061752319336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,1,128,power_law_1.2,0.734997787475586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,1,128,power_law_1.2,0.04849023818969726
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,1,128,power_law_1.2,0.04847104072570801
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,1,128,power_law_1.2,0.9828300476074219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,1,128,power_law_1.2,0.048266239166259765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,1,128,power_law_1.2,0.06393727779388428
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,1,128,power_law_1.2,0.0651097583770752
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,1,128,power_law_1.2,0.06586751937866211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,1,128,power_law_1.2,0.06633471965789794
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,1,128,power_law_1.2,1.5561433410644532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,1,128,power_law_1.2,0.06786816120147705
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,1,128,power_law_1.2,0.06815231800079345
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,1,128,power_law_1.2,0.07132287979125976
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,1,128,power_law_1.2,0.07438079833984375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,1,128,power_law_1.2,0.08122752189636231
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,1,128,power_law_1.2,0.08324352264404297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,1,128,power_law_1.2,2.2442419433593748
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,1,128,power_law_1.2,0.08606719970703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,1,128,power_law_1.2,0.08791551589965821
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,1,128,power_law_1.2,0.09919103622436523
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,1,128,power_law_1.2,0.11760128021240235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,1,128,power_law_1.2,0.16314880371093748
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,1,128,power_law_1.2,0.1956505584716797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,1,128,power_law_1.2,0.07223680019378662
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,1,128,power_law_1.2,0.28729215621948245
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,1,128,power_law_1.2,3.2326577758789066
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,1,128,power_law_1.2,0.462165756225586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,2,64,balanced,0.05051263809204102
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,1,128,power_law_1.2,0.6443647766113282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,2,64,balanced,0.050366721153259277
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,2,64,balanced,0.050170879364013675
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,2,64,balanced,0.05037568092346192
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,1,128,power_law_1.2,0.8601216125488282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,2,64,balanced,0.06307072162628173
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,2,64,balanced,0.07467135906219483
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,1,128,power_law_1.2,0.06898816108703613
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,2,64,balanced,0.12507391929626466
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,2,64,balanced,0.12537599563598634
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,2,64,balanced,0.12628992080688478
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,1,128,power_law_1.2,4.433498840332031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,2,64,balanced,0.1262476825714111
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,2,64,balanced,0.127391996383667
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,1,128,power_law_1.2,1.3591322326660156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,2,64,balanced,0.1280793571472168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,2,64,balanced,0.12902015686035156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,2,64,balanced,0.1352524757385254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,2,64,balanced,0.13805952072143554
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,2,64,balanced,0.14570879936218262
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,2,64,balanced,0.14662272453308106
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,2,64,balanced,0.1516748809814453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,2,64,balanced,0.1621388816833496
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,2,64,balanced,0.18144384384155274
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,1,128,power_law_1.2,1.8065689086914063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,2,64,balanced,0.19190784454345705
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,2,64,balanced,0.2356518363952637
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,2,64,balanced,0.2865958404541016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,2,64,balanced,0.031957759857177734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,2,64,balanced,0.03192831993103028
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,2,64,balanced,0.032143359184265134
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,2,64,balanced,0.03200256109237671
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,2,64,balanced,0.03832448005676269
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,2,64,balanced,0.047895040512084965
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,2,64,balanced,0.06476287841796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,2,64,balanced,0.13684736251831056
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,2,64,balanced,0.06336383819580078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,2,64,balanced,0.1427519989013672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,2,64,balanced,0.06679168224334717
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,2,64,balanced,0.0646131181716919
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,2,64,balanced,0.06583295822143555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,2,64,balanced,0.06686079978942872
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,2,64,balanced,0.0724185609817505
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,2,64,balanced,0.06896512031555176
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,2,64,balanced,0.06942848205566407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,2,64,balanced,0.07133696079254151
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,2,64,balanced,0.07475840091705323
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,2,64,balanced,0.08100095748901368
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,2,64,balanced,0.07581823825836181
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,2,64,balanced,0.08156543731689453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,2,64,balanced,0.08860544204711915
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,32,4,power_law_1.2,0.801585922241211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,2,64,balanced,0.09992832183837891
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,2,64,balanced,0.11362815856933595
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,2,64,balanced,0.14317312240600585
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,2,64,balanced,0.5250803375244141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,2,64,balanced,0.17026176452636718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,2,64,balanced,0.22347904205322267
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,2,64,power_law_1.01,0.07532671928405762
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,2,64,balanced,0.28034048080444335
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,2,64,power_law_1.01,0.07484799861907959
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,2,64,power_law_1.01,0.07517055988311767
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,2,64,power_law_1.01,0.07558911800384521
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,2,64,power_law_1.01,0.12252032279968261
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,2,64,power_law_1.01,0.12704383850097656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,2,64,power_law_1.01,0.12510080337524415
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,2,64,power_law_1.01,0.12537088394165039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,2,64,power_law_1.01,0.13150848388671874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,2,64,power_law_1.01,0.13731200218200684
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,2,64,power_law_1.01,0.13862272262573244
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,2,64,power_law_1.01,0.14478848457336427
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,2,64,balanced,0.4450764846801758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,2,64,power_law_1.01,0.14851712226867675
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,2,64,power_law_1.01,0.1473408031463623
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,2,64,power_law_1.01,0.15437567710876465
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,2,64,power_law_1.01,0.1764031982421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,2,64,power_law_1.01,0.19638143539428712
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,2,64,power_law_1.01,0.23833087921142576
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,2,64,power_law_1.01,0.31124479293823243
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,2,64,power_law_1.01,0.121594877243042
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,2,64,power_law_1.01,0.4421017456054687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,2,64,power_law_1.01,0.1302233600616455
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,2,64,power_law_1.01,0.5444556808471679
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,2,64,power_law_1.01,0.04735743999481201
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,2,64,power_law_1.01,0.04707712173461914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,2,64,power_law_1.01,0.048567042350769044
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,2,64,power_law_1.01,0.7125823974609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,2,64,power_law_1.01,0.04771967887878418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,2,64,power_law_1.01,0.06154623985290527
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,2,64,power_law_1.01,0.06304512023925782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,2,64,power_law_1.01,0.06751872062683105
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,2,64,power_law_1.01,0.06512767791748046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,2,64,power_law_1.01,0.06799744129180908
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,2,64,power_law_1.01,1.0335692596435546
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,2,64,power_law_1.01,0.06835072040557862
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,2,64,power_law_1.01,0.0693670415878296
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,2,64,power_law_1.01,0.07069824218750001
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,2,64,power_law_1.01,0.07483776092529297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,2,64,power_law_1.01,0.07800576210021973
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,2,64,power_law_1.01,0.08229760169982911
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,2,64,power_law_1.01,0.08071552276611328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,2,64,power_law_1.01,0.08315008163452149
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,2,64,power_law_1.01,0.08968576431274414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,2,64,power_law_1.01,0.10206591606140136
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,2,64,power_law_1.01,0.133372163772583
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,2,64,power_law_1.01,0.1561087989807129
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,2,64,power_law_1.01,0.2166886329650879
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,2,64,power_law_1.01,1.8522636413574218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,2,64,power_law_1.01,0.39033344268798825
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,2,64,power_law_1.2,0.0751475191116333
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,2,64,power_law_1.01,0.5192230224609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,2,64,power_law_1.2,0.07492224216461182
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,2,64,power_law_1.2,0.07589119911193848
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,2,64,power_law_1.01,0.7210176086425781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,2,64,power_law_1.2,0.1205183982849121
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,2,64,power_law_1.2,0.12338687896728515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,2,64,power_law_1.01,1.508089599609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,2,64,power_law_1.01,0.9908799743652343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,2,64,power_law_1.2,0.12814592361450194
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,2,64,power_law_1.2,0.12799231529235838
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,2,64,power_law_1.2,0.13346303939819334
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,2,64,power_law_1.01,0.28121856689453123
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,2,64,power_law_1.2,0.14159104347229004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,2,64,power_law_1.2,0.14294272422790527
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,2,64,power_law_1.2,0.14562560081481934
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,2,64,power_law_1.2,0.15007871627807617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,2,64,power_law_1.2,0.15124480247497557
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,2,64,power_law_1.2,0.162476806640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,2,64,power_law_1.2,0.19203199386596678
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,2,64,power_law_1.2,0.07125376224517822
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,2,64,power_law_1.2,0.22012672424316407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,2,64,power_law_1.2,0.29686527252197265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,2,64,power_law_1.2,0.12755583763122558
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,2,64,power_law_1.2,0.3690496063232422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,2,64,power_law_1.2,0.5147776031494141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,2,64,power_law_1.2,0.12595071792602539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,2,64,power_law_1.2,0.7127718353271485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,2,64,power_law_1.2,0.047837438583374026
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,2,64,power_law_1.2,0.047066879272460935
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,2,64,power_law_1.2,0.04800127983093262
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,2,64,power_law_1.2,0.06123263835906982
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,2,64,power_law_1.2,1.126275863647461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,2,64,power_law_1.2,0.061121277809143074
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,2,64,power_law_1.2,1.2630912017822264
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,2,64,power_law_1.2,0.06385536193847656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,2,64,power_law_1.2,0.06601088047027588
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,2,64,power_law_1.2,0.0658355188369751
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,2,64,power_law_1.2,0.06798975944519044
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,2,64,power_law_1.2,0.0670847988128662
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,2,64,power_law_1.2,0.07005695819854736
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,2,64,power_law_1.2,0.07348351955413819
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,2,64,power_law_1.2,0.07760255813598634
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,2,64,power_law_1.2,0.08254079818725586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,2,64,power_law_1.2,0.08086784362792969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,2,64,power_law_1.2,0.08192768096923828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,2,64,power_law_1.2,0.08722944259643554
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,2,64,power_law_1.2,0.0995366382598877
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,2,64,power_law_1.2,0.11611647605895996
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,2,64,power_law_1.2,0.15356672286987305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,2,64,power_law_1.2,0.1673382377624512
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,2,64,power_law_1.2,2.289281311035156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,2,64,power_law_1.2,0.2925913619995117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,2,64,power_law_1.2,0.33428096771240234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,4,32,balanced,0.03139967918395996
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,2,64,power_law_1.2,0.5019404983520508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,4,32,balanced,0.03161600112915039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,4,32,balanced,0.054088959693908686
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,2,64,power_law_1.2,0.7026150512695313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,4,32,balanced,0.06184576034545899
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,4,32,balanced,0.07685120105743408
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,4,32,balanced,0.1251801586151123
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,2,64,power_law_1.2,3.109652404785156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,4,32,balanced,0.12735103607177733
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,4,32,balanced,0.12846976280212402
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,2,64,power_law_1.2,1.1360729980468751
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,4,32,balanced,0.13319680213928223
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,4,32,balanced,0.14317055702209475
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,4,32,balanced,0.13454208374023438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,4,32,balanced,0.14078847885131834
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,4,32,balanced,0.14616064071655271
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,4,32,balanced,0.03054336071014404
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,4,32,balanced,0.13909503936767578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,4,32,balanced,0.14614144325256348
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,4,32,balanced,0.1543680000305176
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,4,32,balanced,0.16776960372924804
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,4,32,balanced,0.1784307289123535
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,4,32,balanced,0.2067942428588867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,4,32,balanced,0.22840831756591795
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,4,32,balanced,0.2869388771057129
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,4,32,balanced,0.12478848457336426
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,4,32,balanced,0.030817279815673827
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,4,32,balanced,0.359420166015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,4,32,balanced,0.03056256055831909
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,4,32,balanced,0.030600960254669192
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,4,32,balanced,0.03152767896652221
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,4,32,balanced,0.03795968055725098
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,4,32,balanced,0.05111680030822754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,4,32,balanced,0.06770944118499755
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,4,32,balanced,0.06912767887115479
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,4,32,balanced,0.0697983980178833
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,4,32,balanced,0.13782655715942382
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,4,32,balanced,0.07000319957733155
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,4,32,balanced,0.07172224044799805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,4,32,balanced,0.07332223892211914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,4,32,balanced,0.07524608135223389
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,4,32,balanced,0.07795199871063233
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,2,64,power_law_1.2,1.622758331298828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,4,32,balanced,0.0810483169555664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,4,32,balanced,0.07494016170501709
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,4,32,balanced,0.081342716217041
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,4,32,balanced,0.0866483211517334
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,4,32,balanced,0.10990464210510253
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,4,32,balanced,0.098853120803833
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,4,32,balanced,0.1555788803100586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,4,32,balanced,0.20779136657714842
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,4,32,balanced,0.5535641479492187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,4,32,balanced,0.3442559814453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,4,32,balanced,0.6842739105224609
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,4,32,power_law_1.01,0.07817344188690185
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,4,32,balanced,0.44035583496093744
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,4,32,balanced,0.07158143997192383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,4,32,power_law_1.01,0.07743487834930421
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,4,32,power_law_1.01,0.07897727966308594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,4,32,power_law_1.01,0.12010111808776855
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,4,32,power_law_1.01,0.11775872230529785
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,4,32,power_law_1.01,0.12317440032958984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,4,32,power_law_1.01,0.1253388786315918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,4,32,power_law_1.01,0.12953472137451172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,4,32,power_law_1.01,0.12719231605529785
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,4,32,power_law_1.01,0.13459199905395508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,4,32,power_law_1.01,0.13637503623962402
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,4,32,power_law_1.01,0.1383846378326416
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,4,32,power_law_1.01,0.14519807815551758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,4,32,power_law_1.01,0.14187520027160644
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,4,32,balanced,0.24952959060668944
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,4,32,power_law_1.01,0.14644351959228516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,4,32,power_law_1.01,0.15432319641113282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,4,32,power_law_1.01,0.1695142364501953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,4,32,power_law_1.01,0.18678144454956053
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,4,32,power_law_1.01,0.21939199447631835
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,4,32,power_law_1.01,0.2642995262145996
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,4,32,power_law_1.01,0.3437516784667969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,4,32,power_law_1.01,0.42385406494140626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,4,32,balanced,0.1328166389465332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,4,32,power_law_1.01,0.04979072093963623
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,4,32,power_law_1.01,0.6562777709960937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,4,32,power_law_1.01,0.04990464210510254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,4,32,power_law_1.01,0.769027862548828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,4,32,power_law_1.01,0.04998015880584717
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,4,32,power_law_1.01,0.051376638412475584
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,4,32,power_law_1.01,0.07809152126312255
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,4,32,power_law_1.01,0.06625664234161377
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,4,32,power_law_1.01,0.0666534423828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,4,32,power_law_1.01,1.1711065673828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,4,32,power_law_1.01,0.07077504158020019
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,4,32,power_law_1.01,0.07081344127655029
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,4,32,power_law_1.01,0.07341440200805664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,4,32,power_law_1.01,0.07519487857818603
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,4,32,power_law_1.01,0.07763584136962891
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,4,32,power_law_1.01,0.08429183959960937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,4,32,power_law_1.01,0.07883776187896728
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,4,32,power_law_1.01,1.578005828857422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,4,32,power_law_1.01,0.07951104164123535
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,4,32,power_law_1.01,0.08435968399047852
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,4,32,power_law_1.01,0.09730303764343262
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,4,32,power_law_1.01,0.10478079795837403
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,4,32,power_law_1.01,0.1302783966064453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,4,32,power_law_1.01,0.16156543731689452
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,4,32,power_law_1.01,0.2145702362060547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,4,32,power_law_1.01,0.26566911697387696
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,4,32,power_law_1.01,0.06767360210418702
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,4,32,power_law_1.01,0.4071257781982422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,4,32,power_law_1.01,0.06975232124328613
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,4,32,power_law_1.2,0.07792384147644042
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,4,32,power_law_1.01,0.4790476989746094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,4,32,power_law_1.2,0.0778444814682007
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,4,32,power_law_1.2,0.07650944232940673
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,4,32,power_law_1.2,0.11939968109130858
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,4,32,power_law_1.2,0.11670144081115723
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,4,32,power_law_1.2,0.1221247959136963
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,4,32,power_law_1.2,0.12452223777770996
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,4,32,power_law_1.01,0.7927859497070313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,4,32,power_law_1.2,0.12486144065856934
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,4,32,power_law_1.2,0.12816639900207522
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,4,32,power_law_1.2,0.13414400100708007
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,4,32,power_law_1.2,0.13664511680603025
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,4,32,power_law_1.2,0.14272512435913084
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,4,32,power_law_1.01,0.961670379638672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,4,32,power_law_1.2,0.14609791755676269
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,4,32,power_law_1.2,0.1464857578277588
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,4,32,power_law_1.2,0.15274880409240724
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,4,32,power_law_1.2,0.15850624084472656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,4,32,power_law_1.2,0.17542144775390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,4,32,power_law_1.2,0.20054399490356448
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,4,32,power_law_1.2,0.2396031951904297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,4,32,power_law_1.2,0.0776793622970581
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,4,32,power_law_1.2,0.3396748733520508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,4,32,power_law_1.2,0.402790412902832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,4,32,power_law_1.2,0.04977151870727539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,4,32,power_law_1.2,0.050007038116455085
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,4,32,power_law_1.2,0.05016704082489014
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,4,32,power_law_1.2,0.04980991840362549
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,4,32,power_law_1.2,0.909749755859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,4,32,power_law_1.2,0.06528768062591553
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,4,32,power_law_1.2,0.9643097686767579
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,4,32,power_law_1.2,0.0680191993713379
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,4,32,power_law_1.2,0.07006720066070557
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,4,32,power_law_1.2,0.06940544128417969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,4,32,power_law_1.2,0.07077760219573974
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,4,32,power_law_1.2,0.0739135980606079
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,4,32,power_law_1.2,0.0765235185623169
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,4,32,power_law_1.2,0.07966847896575928
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,4,32,power_law_1.2,0.5415436935424804
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,4,32,power_law_1.2,0.08308735847473145
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,4,32,power_law_1.2,0.08705280303955079
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,4,32,power_law_1.2,0.10318592071533203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,4,32,power_law_1.2,0.11961088180541993
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,4,32,power_law_1.2,0.0651263999938965
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,4,32,power_law_1.2,0.15000191688537598
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,4,32,power_law_1.2,2.1494668579101566
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,4,32,power_law_1.2,0.17220352172851563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,4,32,power_law_1.2,0.2314227294921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,4,32,power_law_1.2,0.3121881675720215
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,4,32,power_law_1.2,0.08231040000915527
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,8,16,balanced,0.030337278842926023
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,4,32,power_law_1.2,0.08029567718505859
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,8,16,balanced,0.030406401157379147
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,4,32,power_law_1.2,1.7273727416992188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,8,16,balanced,0.03166079998016357
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,8,16,balanced,0.05425151824951172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,4,32,power_law_1.2,0.607399673461914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,8,16,balanced,0.06422143936157226
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,8,16,balanced,0.08771967887878418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,8,16,balanced,0.1297267246246338
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,8,16,balanced,0.13370752334594727
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,8,16,balanced,0.13635199546813964
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,8,16,balanced,0.1403980827331543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,8,16,balanced,0.1431999969482422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,8,16,balanced,0.14026368141174317
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,4,32,power_law_1.2,0.5259596633911132
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,4,32,power_law_1.2,1.5263462829589844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,8,16,balanced,0.14339712142944336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,8,16,balanced,0.15023743629455566
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,8,16,balanced,0.13000191688537596
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,8,16,balanced,0.16917503356933594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,8,16,balanced,0.16043136596679688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,8,16,balanced,0.13283200263977052
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,8,16,balanced,0.19267711639404297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,8,16,balanced,0.21466495513916017
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,4,32,power_law_1.2,1.060479965209961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,8,16,balanced,0.26270719528198244
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,8,16,balanced,0.30496768951416015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,8,16,balanced,0.1367743968963623
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,8,16,balanced,0.03045248031616211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,8,16,balanced,0.4006643295288086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,8,16,balanced,0.03054336071014404
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,8,16,balanced,0.03147264003753662
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,8,16,balanced,0.03277184009552002
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,8,16,balanced,0.043322877883911134
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,8,16,balanced,0.05837696075439454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,8,16,balanced,0.08484352111816407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,8,16,balanced,0.07849855899810791
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,8,16,balanced,0.07908224105834961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,8,16,balanced,0.8011583709716797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,8,16,balanced,0.08236031532287598
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,8,16,balanced,0.0848755168914795
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,8,16,balanced,0.08819071769714355
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,8,16,balanced,0.08109439849853516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,8,16,balanced,0.08678527832031249
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,8,16,balanced,0.09239680290222167
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,8,16,balanced,0.10163071632385254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,8,16,balanced,0.11229056358337401
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,8,16,balanced,0.13334272384643556
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,8,16,balanced,0.15607423782348634
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,8,16,balanced,0.5203225708007813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,8,16,balanced,0.23729536056518558
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,8,16,balanced,0.07921279907226562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,8,16,balanced,0.3337343978881836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,8,16,balanced,0.41367168426513673
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,8,16,power_law_1.01,0.06336512088775634
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,8,16,balanced,0.08481280326843262
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,8,16,power_law_1.01,0.0756608009338379
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,8,16,balanced,0.591153907775879
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,8,16,power_law_1.01,0.07534848213195801
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,8,16,power_law_1.01,0.07769472122192382
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,8,16,balanced,0.7728409576416015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,8,16,power_law_1.01,0.08014335632324218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,8,16,power_law_1.01,0.1197657585144043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,8,16,power_law_1.01,0.12366975784301756
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,8,16,power_law_1.01,0.1303872013092041
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,8,16,power_law_1.01,0.1360217571258545
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,8,16,power_law_1.01,0.1317427158355713
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,8,16,power_law_1.01,0.1377996826171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,8,16,power_law_1.01,0.14369791984558106
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,8,16,power_law_1.01,0.13744383811950683
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,8,16,power_law_1.01,0.14148608207702637
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,8,16,power_law_1.01,0.14719615936279298
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,8,16,power_law_1.01,0.17229055404663085
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,8,16,power_law_1.01,0.18236928939819336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,8,16,balanced,0.9956403350830078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,8,16,power_law_1.01,0.21210880279541017
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,8,16,power_law_1.01,0.26475008010864254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,8,16,power_law_1.01,0.33709182739257815
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,8,16,balanced,0.1973632049560547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,8,16,power_law_1.01,0.4558694458007812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,8,16,power_law_1.01,0.6192358398437501
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,8,16,power_law_1.01,0.0404531192779541
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,8,16,power_law_1.01,0.05397503852844239
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,8,16,power_law_1.01,0.15393792152404787
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,8,16,power_law_1.01,0.8842291259765626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,8,16,power_law_1.01,0.05357439994812012
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,8,16,power_law_1.01,0.055127038955688476
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,8,16,power_law_1.01,0.05513855934143066
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,8,16,power_law_1.01,0.07183487892150879
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,8,16,power_law_1.01,1.1222207641601563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,8,16,power_law_1.01,0.0789363193511963
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,8,16,power_law_1.01,0.07516672134399413
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,8,16,power_law_1.01,0.07627903938293457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,8,16,power_law_1.01,0.07861631870269775
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,8,16,power_law_1.01,0.0830720043182373
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,8,16,power_law_1.01,0.08888704299926757
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,8,16,power_law_1.01,0.1256460762023926
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,8,16,power_law_1.01,0.08913151741027832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,8,16,power_law_1.01,0.0943295955657959
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,8,16,power_law_1.01,0.1092684841156006
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,8,16,power_law_1.01,0.12351743698120117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,8,16,power_law_1.01,1.5749440002441406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,8,16,power_law_1.01,0.1685887908935547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,8,16,power_law_1.01,0.1960972785949707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,8,16,power_law_1.01,0.25395328521728516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,8,16,power_law_1.01,0.269036808013916
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,8,16,power_law_1.2,0.06330624103546142
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,8,16,power_law_1.01,0.4510092926025391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,8,16,power_law_1.01,0.08548735618591309
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,8,16,power_law_1.01,0.08356096267700194
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,8,16,power_law_1.2,0.07552896022796632
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,8,16,power_law_1.2,0.07385727882385254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,8,16,power_law_1.2,0.07652607917785645
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,8,16,power_law_1.01,0.8607628631591797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,8,16,power_law_1.2,0.07879295825958252
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,8,16,power_law_1.2,0.11601152420043945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,8,16,power_law_1.2,0.12331647872924804
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,8,16,power_law_1.2,0.12957311630249024
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,8,16,power_law_1.2,0.1291097640991211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,8,16,power_law_1.01,1.1847001647949218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,8,16,power_law_1.2,0.13228416442871094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,8,16,power_law_1.2,0.13768832206726073
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,8,16,power_law_1.2,0.14030847549438477
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,8,16,power_law_1.2,0.1446566390991211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,8,16,power_law_1.2,0.15065343856811525
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,8,16,power_law_1.2,0.1468825626373291
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,8,16,power_law_1.2,0.14911871910095215
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,8,16,power_law_1.2,0.1575500774383545
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,8,16,power_law_1.2,0.17628543853759765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,8,16,power_law_1.2,0.19963775634765626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,8,16,power_law_1.2,0.24152063369750976
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,8,16,power_law_1.2,0.279289608001709
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,8,16,power_law_1.2,0.4014080047607422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,8,16,power_law_1.2,0.504273910522461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,8,16,power_law_1.2,0.6812850952148437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,8,16,power_law_1.01,0.08356608390808105
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,8,16,power_law_1.2,0.05401343822479247
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,8,16,power_law_1.2,0.051904001235961915
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,8,16,power_law_1.2,1.053350372314453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,8,16,power_law_1.2,0.05434879779815673
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,8,16,power_law_1.2,0.056421117782592775
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,8,16,power_law_1.2,0.07032703876495361
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,8,16,power_law_1.2,0.0760319995880127
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,8,16,power_law_1.2,1.5366937255859374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,8,16,power_law_1.2,0.07560959815979004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,8,16,power_law_1.2,0.07834496021270751
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,8,16,power_law_1.2,0.07939839839935303
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,8,16,power_law_1.2,0.08277503967285156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,8,16,power_law_1.2,0.04061567783355713
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,8,16,power_law_1.01,0.5839052963256836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,8,16,power_law_1.2,0.08757887840270996
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,8,16,power_law_1.2,0.08766847610473633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,8,16,power_law_1.2,0.09005824089050293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,8,16,power_law_1.2,1.9005836486816405
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,8,16,power_law_1.2,0.13007871627807616
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,8,16,power_law_1.2,0.11549311637878419
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,8,16,power_law_1.2,0.17073919296264647
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,8,16,power_law_1.2,0.21080575942993166
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,8,16,power_law_1.2,0.27630847930908203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,8,16,power_law_1.2,0.3742067337036133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,8,16,power_law_1.2,0.5544704055786133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,16,8,balanced,0.030716159343719483
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,8,16,power_law_1.2,0.08650752067565917
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,8,16,power_law_1.2,0.08451583862304687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,16,8,balanced,0.04418176174163818
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,16,8,balanced,0.031650559902191164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,16,8,balanced,0.05985536098480224
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,16,8,balanced,0.07145472049713135
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,16,8,balanced,0.0890060806274414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,16,8,balanced,0.137639684677124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,16,8,balanced,0.14102911949157715
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,16,8,balanced,0.14463871955871582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,8,16,power_law_1.2,1.0705677032470704
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,16,8,balanced,0.13811200141906738
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,16,8,balanced,0.14468864440917967
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,16,8,balanced,0.14955519676208495
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,16,8,balanced,0.15648768424987794
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,16,8,balanced,0.16091007232666016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,16,8,balanced,0.16900352478027342
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,8,16,power_law_1.2,1.4314790344238282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,16,8,balanced,0.18998655319213867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,8,16,power_law_1.2,0.7264793395996094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,16,8,balanced,0.2087539291381836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,16,8,balanced,0.25001983642578124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,8,16,power_law_1.2,0.10040960311889649
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,16,8,balanced,0.13393280029296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,16,8,balanced,0.2918041610717773
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,16,8,balanced,0.3821478271484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,16,8,balanced,0.030938880443572996
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,16,8,balanced,0.03132800102233887
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,16,8,balanced,0.14063615798950196
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,16,8,balanced,0.03436543941497803
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,16,8,balanced,0.03896703958511353
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,16,8,balanced,0.05414400100708008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,16,8,balanced,0.0708569622039795
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,16,8,balanced,0.09400959968566894
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,16,8,balanced,0.8313983917236328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,16,8,balanced,0.09689215660095216
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,16,8,balanced,0.10032896041870117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,16,8,balanced,0.10079744338989258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,16,8,balanced,0.09835519790649414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,16,8,balanced,0.09537792205810547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,16,8,balanced,0.10044159889221191
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,16,8,balanced,0.10607744216918946
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,16,8,balanced,1.2907571411132812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,16,8,balanced,0.11416192054748535
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,16,8,balanced,0.1245900821685791
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,16,8,balanced,0.14339712142944336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,16,8,balanced,0.46164993286132816
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,16,8,balanced,0.16523263931274415
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,16,8,balanced,0.20490623474121095
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,16,8,balanced,0.6339315032958984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,16,8,balanced,1.6625970458984376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,16,8,balanced,0.3252159881591797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,16,8,balanced,0.4009574508666992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,16,8,balanced,0.5893657684326172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,16,8,power_law_1.01,0.05590911865234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,16,8,balanced,0.10892160415649414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,16,8,balanced,0.7437606048583985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,16,8,power_law_1.01,0.06477312088012696
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,16,8,power_law_1.01,0.06693632125854491
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,16,8,power_law_1.01,0.0742796802520752
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,16,8,balanced,1.0889574432373048
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,16,8,power_law_1.01,0.08160384178161621
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,16,8,power_law_1.01,0.12469759941101075
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,16,8,power_law_1.01,0.12975104331970216
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,16,8,power_law_1.01,0.1370316791534424
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,16,8,power_law_1.01,0.14264063835144042
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,16,8,power_law_1.01,0.13872768402099608
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,16,8,balanced,1.4369151306152343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,16,8,power_law_1.01,0.13702143669128417
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,16,8,power_law_1.01,0.14153216361999513
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,16,8,power_law_1.01,0.14400256156921387
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,16,8,power_law_1.01,0.15246463775634767
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,16,8,power_law_1.01,0.1602662467956543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,16,8,power_law_1.01,0.17553792953491212
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,16,8,power_law_1.01,0.1939276885986328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,16,8,power_law_1.01,0.07880320072174071
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,16,8,power_law_1.01,0.2372224044799805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,16,8,power_law_1.01,0.2762188720703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,16,8,power_law_1.01,0.3367116928100586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,16,8,power_law_1.01,0.13214464187622071
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,16,8,power_law_1.01,0.44729854583740236
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,16,8,power_law_1.01,0.035699200630187986
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,16,8,power_law_1.01,0.6288371276855469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,16,8,power_law_1.01,0.047802882194519045
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,16,8,power_law_1.01,0.055925760269165035
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,16,8,power_law_1.01,0.8130303955078124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,16,8,power_law_1.01,0.060550398826599114
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,16,8,power_law_1.01,0.06744959831237793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,16,8,balanced,0.24549760818481445
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,16,8,power_law_1.01,0.08826751708984375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,16,8,power_law_1.01,0.08958208084106445
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,16,8,power_law_1.01,0.09112447738647461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,16,8,power_law_1.01,0.09422975540161133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,16,8,power_law_1.01,0.0985318374633789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,16,8,power_law_1.01,1.2435187530517577
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,16,8,power_law_1.01,0.09736703872680665
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,16,8,power_law_1.01,0.09336832046508789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,16,8,power_law_1.01,0.09849599838256837
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,16,8,power_law_1.01,0.10637439727783202
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,16,8,power_law_1.01,0.13470080375671387
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,16,8,power_law_1.01,0.15591423988342284
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,16,8,power_law_1.01,1.6113714599609374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,16,8,power_law_1.01,0.18128896713256837
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,16,8,power_law_1.01,0.23229055404663085
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,16,8,power_law_1.01,0.05269504070281983
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,16,8,power_law_1.01,0.3121139144897461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,16,8,power_law_1.01,0.41139457702636717
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,16,8,power_law_1.2,0.0551859188079834
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,16,8,power_law_1.01,0.527438087463379
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,16,8,power_law_1.2,0.06397823810577392
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,16,8,power_law_1.01,0.6715417480468749
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,16,8,power_law_1.2,0.06392704010009766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,16,8,power_law_1.01,0.11339391708374022
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,16,8,power_law_1.2,0.07104256153106689
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,16,8,power_law_1.2,0.07881472110748292
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,16,8,power_law_1.01,1.0934207916259766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,16,8,power_law_1.2,0.1307699203491211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,16,8,power_law_1.2,0.13286656379699707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,16,8,power_law_1.2,0.1358950424194336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,16,8,power_law_1.01,1.4238182067871095
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,16,8,power_law_1.2,0.138088960647583
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,16,8,power_law_1.2,0.13979519844055174
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,16,8,power_law_1.2,0.1429248046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,16,8,power_law_1.2,0.15343232154846193
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,16,8,power_law_1.2,0.1589235210418701
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,16,8,power_law_1.2,0.15187456130981444
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,16,8,power_law_1.01,0.10084351539611816
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,16,8,power_law_1.2,0.18395904541015623
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,16,8,power_law_1.2,0.07749760150909424
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,16,8,power_law_1.2,0.2013158416748047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,16,8,power_law_1.2,0.12654848098754884
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,16,8,power_law_1.2,0.24338943481445313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,16,8,power_law_1.2,0.2955558395385742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,16,8,power_law_1.2,0.36474750518798826
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,16,8,power_law_1.2,0.4724646377563476
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,16,8,power_law_1.2,0.03552383899688721
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,16,8,power_law_1.2,0.14081536293029787
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,16,8,power_law_1.2,0.04726272106170654
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,16,8,power_law_1.2,0.04578944206237793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,16,8,power_law_1.2,0.05507455825805664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,16,8,power_law_1.2,0.06222208023071289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,16,8,power_law_1.2,0.06688255786895751
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,16,8,power_law_1.2,0.821006088256836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,16,8,power_law_1.2,0.08791935920715332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,16,8,power_law_1.2,0.09011839866638184
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,16,8,power_law_1.2,0.09163392066955567
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,16,8,power_law_1.2,0.09600255966186524
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,16,8,power_law_1.2,0.09900671958923339
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,16,8,power_law_1.2,0.09464703559875488
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,16,8,power_law_1.2,1.316259765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,16,8,power_law_1.2,0.09916799545288087
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,16,8,power_law_1.2,0.1054860782623291
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,16,8,power_law_1.2,0.10829183578491211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,16,8,power_law_1.2,0.11277055740356445
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,16,8,power_law_1.2,0.13941503524780274
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,16,8,power_law_1.2,0.16026367187499999
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,16,8,power_law_1.2,1.8603417968750002
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,16,8,power_law_1.2,0.7535987091064453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,16,8,power_law_1.2,0.20591360092163086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,16,8,power_law_1.2,0.2535987281799316
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,16,8,power_law_1.2,0.30824960708618165
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,16,8,power_law_1.2,0.4149951934814453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,32,4,balanced,0.03020031929016113
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,32,4,balanced,0.051151361465454105
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,16,8,power_law_1.2,0.6136115264892579
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,16,8,power_law_1.2,0.6774822235107422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,32,4,balanced,0.06548736095428467
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,32,4,balanced,0.07412352085113524
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,32,4,balanced,0.09668992042541505
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,32,4,balanced,0.09514880180358887
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,32,4,balanced,0.09833727836608887
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,32,4,balanced,0.13751296043395994
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,32,4,balanced,0.13962112426757814
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,16,8,power_law_1.2,1.1951718139648437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,32,4,balanced,0.14311424255371094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,32,4,balanced,0.1456383991241455
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,32,4,balanced,0.1489510440826416
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,32,4,balanced,0.1541222381591797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,16,8,power_law_1.2,0.09908608436584473
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,32,4,balanced,0.1587724781036377
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,16,8,power_law_1.2,1.5057894897460937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,32,4,balanced,0.1707276725769043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,32,4,balanced,0.031823360919952394
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,32,4,balanced,0.18998783111572265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,32,4,balanced,0.20924160003662112
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,32,4,balanced,0.2505907249450684
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,32,4,balanced,0.2911078453063965
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,32,4,balanced,0.38366592407226563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,32,4,balanced,0.46562686920166013
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,32,4,balanced,0.03014656066894531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,32,4,balanced,0.03229055881500244
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,32,4,balanced,0.6379699325561523
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,32,4,balanced,0.03624703884124756
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,32,4,balanced,0.046370558738708496
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,32,4,balanced,0.06195968151092529
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,32,4,balanced,0.09204863548278809
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,32,4,balanced,0.0926195240020752
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,32,4,balanced,0.16257280349731446
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,32,4,balanced,0.12338175773620605
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,32,4,balanced,0.12299263954162598
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,32,4,balanced,0.1290700817108154
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,32,4,balanced,0.13250432014465333
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,32,4,balanced,1.2865420532226561
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,32,4,balanced,0.1357414436340332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,32,4,balanced,0.14003840446472168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,32,4,balanced,0.14423295974731445
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,32,4,balanced,0.153570556640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,32,4,balanced,0.17313024520874024
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,32,4,balanced,0.19278976440429688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,32,4,balanced,0.23325311660766604
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,32,4,balanced,1.73417724609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,32,4,balanced,0.2674009513854981
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,32,4,balanced,0.829534683227539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,32,4,balanced,0.34514560699462893
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,32,4,balanced,0.08919936180114746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,32,4,balanced,0.41970176696777345
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,32,4,balanced,0.12674431800842284
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,32,4,power_law_1.01,0.03311615943908691
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,32,4,power_law_1.01,0.05778816223144532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,32,4,power_law_1.01,0.06407296180725097
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,32,4,balanced,0.6298880004882813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,32,4,power_law_1.01,0.07025792121887206
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,32,4,power_law_1.01,0.07664127826690674
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,32,4,power_law_1.01,0.0844428825378418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,32,4,balanced,0.7829325103759766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,32,4,power_law_1.01,0.09487487792968749
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,32,4,power_law_1.01,0.09039615631103516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,32,4,power_law_1.01,0.1334284782409668
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,32,4,power_law_1.01,0.13621503829956055
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,32,4,power_law_1.01,0.13870847702026368
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,32,4,power_law_1.01,0.14225024223327637
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,32,4,balanced,1.1843251037597657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,32,4,power_law_1.01,0.14827391624450684
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,32,4,power_law_1.01,0.1533670425415039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,32,4,power_law_1.01,0.15784576416015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,32,4,power_law_1.01,0.17015935897827147
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,32,4,power_law_1.01,0.18889087677001953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,32,4,power_law_1.01,0.2134771156311035
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,32,4,power_law_1.01,0.2555046463012695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,32,4,balanced,1.5468978881835938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,32,4,power_law_1.01,0.29647615432739255
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,32,4,power_law_1.01,0.3823040008544922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,32,4,power_law_1.01,0.03490560054779053
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,32,4,power_law_1.01,0.46359935760498044
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,32,4,power_law_1.01,0.1325004768371582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,32,4,power_law_1.01,0.040945920944213864
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,32,4,power_law_1.01,0.6681177520751953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,32,4,power_law_1.01,0.04958079814910889
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,32,4,power_law_1.01,0.058982400894165045
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,32,4,power_law_1.01,0.8832268524169923
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,32,4,power_law_1.01,0.07720960140228271
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,32,4,power_law_1.01,0.08104063987731933
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,32,4,power_law_1.01,0.08067584037780762
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,32,4,power_law_1.01,0.11078528404235839
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,32,4,power_law_1.01,0.11383935928344727
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,32,4,power_law_1.01,1.2879180908203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,32,4,power_law_1.01,0.11458047866821289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,32,4,power_law_1.01,0.11994751930236816
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,32,4,power_law_1.01,0.12142335891723632
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,32,4,power_law_1.01,0.1286963176727295
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,32,4,power_law_1.01,0.13789567947387696
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,32,4,power_law_1.01,0.14715392112731934
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,32,4,power_law_1.01,0.13392640113830567
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,32,4,power_law_1.01,1.7773735046386718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,32,4,power_law_1.01,0.16909183502197264
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,32,4,power_law_1.01,0.23450880050659179
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,32,4,power_law_1.01,0.2827519989013672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,32,4,power_law_1.01,0.37557632446289063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,32,4,power_law_1.01,0.45622528076171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,32,4,power_law_1.01,0.6723430633544922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,32,4,power_law_1.01,0.8849932861328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,32,4,power_law_1.2,0.05747200012207031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,32,4,power_law_1.01,0.06550271987915039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,32,4,power_law_1.2,0.06111743927001954
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,32,4,power_law_1.2,0.0709990406036377
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,32,4,power_law_1.2,0.07618559837341309
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,32,4,power_law_1.2,0.07655168056488038
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,32,4,power_law_1.01,1.2746598052978517
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,32,4,power_law_1.2,0.09333503723144532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,32,4,power_law_1.2,0.09614848136901856
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,32,4,power_law_1.2,0.1326476764678955
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,32,4,power_law_1.2,0.13262847900390623
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,32,4,power_law_1.2,0.13556608200073242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,32,4,power_law_1.2,0.13893376350402833
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,32,4,power_law_1.2,0.03326848030090332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,32,4,power_law_1.2,0.1427443218231201
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,32,4,power_law_1.2,0.15019007682800295
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,32,4,power_law_1.2,0.15443072319030762
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,32,4,power_law_1.2,0.16006271362304686
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,32,4,power_law_1.2,0.17228544235229493
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,32,4,power_law_1.2,0.19251712799072265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,32,4,power_law_1.01,0.19592063903808593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,32,4,power_law_1.2,0.2569625663757324
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,32,4,power_law_1.2,0.2984614372253418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,32,4,power_law_1.2,0.38913536071777344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,32,4,power_law_1.2,0.4782694244384766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,32,4,power_law_1.2,0.03476095914840698
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,32,4,power_law_1.2,0.716968994140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,32,4,power_law_1.01,1.6893901062011718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,32,4,power_law_1.2,0.04105088233947754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,32,4,power_law_1.2,0.04771455764770508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,32,4,power_law_1.2,0.9416089630126953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,32,4,power_law_1.2,0.057032961845397946
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,32,4,power_law_1.2,0.2126374435424805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,32,4,power_law_1.2,0.07613696098327637
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,32,4,power_law_1.2,0.06183040142059326
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,32,4,power_law_1.2,0.0632806396484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,32,4,power_law_1.2,0.09094143867492675
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,32,4,power_law_1.2,0.10854144096374511
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,32,4,power_law_1.2,0.11449855804443358
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,32,4,power_law_1.2,1.3463320922851563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,32,4,power_law_1.2,0.12201984405517578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,32,4,power_law_1.2,0.12997504234313967
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,32,4,power_law_1.2,0.13385600090026856
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,32,4,power_law_1.2,0.13995008468627929
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,32,4,power_law_1.2,0.1727398490905762
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,32,4,power_law_1.2,1.840326385498047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,32,4,power_law_1.2,0.19944448471069337
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,32,4,power_law_1.2,0.24706560134887695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,32,4,power_law_1.2,0.28942592620849605
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,32,4,power_law_1.2,0.38555648803710935
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,32,4,power_law_1.2,0.11218432426452636
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,32,4,power_law_1.2,0.48635520935058596
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,32,4,power_law_1.2,0.11920000076293946
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,32,4,power_law_1.2,0.6991180419921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,1,128,balanced,0.06153088092803956
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,1,128,balanced,0.06133759975433349
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,32,4,power_law_1.2,0.15251584053039552
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,1,128,balanced,0.06146815776824951
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,1,128,balanced,0.061456642150878905
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,32,4,power_law_1.2,0.91246337890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,1,128,balanced,0.12044544219970703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,1,128,balanced,0.12122367858886718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,1,128,balanced,0.12082816123962403
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,1,128,balanced,0.12164863586425781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,1,128,balanced,0.12153087615966798
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,1,128,balanced,0.12427264213562013
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,1,128,balanced,0.12543999671936035
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,1,128,balanced,0.12885631561279295
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,1,128,balanced,0.1317542362213135
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,1,128,balanced,0.129432315826416
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,1,128,balanced,0.13318143844604494
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,1,128,balanced,0.06122879981994629
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,1,128,balanced,0.1402444839477539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,1,128,balanced,0.14581119537353515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,1,128,balanced,0.14071935653686524
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,1,128,balanced,0.1563251209259033
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,32,4,power_law_1.2,1.7528410339355467
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,1,128,balanced,0.1593280029296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,1,128,balanced,0.19422719955444337
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,1,128,balanced,0.22989696502685547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,1,128,balanced,0.03593983888626099
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,1,128,balanced,0.34118526458740234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,1,128,balanced,0.39161598205566406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,1,128,balanced,0.03555072069168091
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,1,128,balanced,0.03225343942642212
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,1,128,balanced,0.03632256031036377
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,1,128,balanced,0.04444799900054931
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,1,128,balanced,0.06015744209289551
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,1,128,balanced,0.06017920017242432
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,1,128,balanced,0.06046336174011231
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,1,128,balanced,0.06041215896606446
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,1,128,balanced,0.06148223876953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,1,128,balanced,0.0698419189453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,1,128,balanced,0.06177279949188232
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,1,128,balanced,0.06156799793243408
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,1,128,balanced,0.06371327877044677
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,1,128,balanced,0.06318336009979249
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,1,128,balanced,0.06456704139709472
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,1,128,balanced,0.06542208194732665
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,1,128,balanced,0.06739200115203857
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,1,128,balanced,0.0352128005027771
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,1,128,balanced,0.0778547191619873
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,32,4,power_law_1.2,1.3062591552734375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,1,128,balanced,0.07241983890533447
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,1,128,balanced,0.08016256332397462
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,1,128,balanced,0.08768896102905274
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,1,128,balanced,0.10486528396606445
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,1,128,balanced,0.11862015724182129
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,1,128,balanced,0.15088255882263185
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,1,128,balanced,0.18895231246948244
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,1,128,power_law_1.01,0.07050367832183838
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,1,128,power_law_1.01,0.06998144149780273
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,1,128,power_law_1.01,0.06975743770599366
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,1,128,power_law_1.01,0.07038464069366454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,1,128,power_law_1.01,0.12348159790039062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,1,128,power_law_1.01,0.11919360160827637
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,1,128,power_law_1.01,0.12231807708740235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,1,128,power_law_1.01,0.12378879547119141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,1,128,power_law_1.01,0.1264742374420166
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,1,128,power_law_1.01,0.1324019241333008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,1,128,balanced,0.07085567951202393
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,1,128,power_law_1.01,0.13071999549865723
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,1,128,balanced,0.1244262409210205
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,1,128,power_law_1.01,0.1298739242553711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,1,128,power_law_1.01,0.14227328300476075
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,1,128,power_law_1.01,0.1405183982849121
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,1,128,power_law_1.01,0.14829952239990235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,1,128,power_law_1.01,0.15896063804626465
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,1,128,power_law_1.01,0.17808000564575194
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,1,128,power_law_1.01,0.23631359100341798
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,1,128,power_law_1.01,0.3170969581604004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,1,128,power_law_1.01,0.46536705017089847
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,1,128,power_law_1.01,0.5252671813964843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,1,128,power_law_1.01,0.04464511871337891
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,1,128,power_law_1.01,0.04469247817993164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,1,128,power_law_1.01,0.8375949096679687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,1,128,power_law_1.01,0.1404646396636963
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,1,128,power_law_1.01,0.04503039836883545
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,1,128,power_law_1.01,1.0495986938476562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,1,128,power_law_1.01,0.044618239402771
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,1,128,power_law_1.01,0.059953918457031255
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,1,128,power_law_1.01,0.060797438621521
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,1,128,power_law_1.01,0.0618291187286377
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,1,128,power_law_1.01,0.062311677932739264
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,1,128,power_law_1.01,0.06288640022277832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,1,128,power_law_1.01,0.06442495822906494
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,1,128,power_law_1.01,0.06452608108520508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,1,128,power_law_1.01,0.06598144054412843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,1,128,power_law_1.01,0.06952703952789306
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,1,128,power_law_1.01,0.12998656272888182
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,1,128,power_law_1.01,0.07335807800292968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,1,128,power_law_1.01,0.07944575786590577
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,1,128,power_law_1.01,0.08061056137084961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,1,128,power_law_1.01,1.8822962951660156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,1,128,power_law_1.01,0.09349120140075684
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,1,128,power_law_1.01,0.11296128273010253
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,1,128,power_law_1.01,0.13353471755981444
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,1,128,power_law_1.01,0.061514239311218265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,1,128,power_law_1.01,0.1987673568725586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,1,128,power_law_1.01,0.24906368255615235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,1,128,power_law_1.01,2.326379547119141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,1,128,power_law_1.01,0.35643009185791014
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,1,128,power_law_1.2,0.07069568157196045
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,1,128,power_law_1.01,0.45097599029541013
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,1,128,power_law_1.2,0.07021567821502686
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,1,128,power_law_1.2,0.07098624229431152
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,1,128,power_law_1.2,0.12024831771850586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,1,128,power_law_1.01,0.07393280029296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,1,128,power_law_1.2,0.1231820774078369
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,1,128,power_law_1.01,0.7473522949218749
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,1,128,power_law_1.2,0.12654208183288573
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,1,128,power_law_1.2,0.12721535682678223
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,1,128,power_law_1.2,0.12912768363952637
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,1,128,power_law_1.2,0.1304255962371826
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,1,128,power_law_1.2,0.13669119834899904
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,1,128,power_law_1.2,0.13012607574462892
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,1,128,power_law_1.01,1.120865249633789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,1,128,power_law_1.2,0.14383744239807128
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,1,128,power_law_1.2,0.1472422409057617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,1,128,power_law_1.2,0.14907520294189452
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,1,128,power_law_1.2,0.15117695808410644
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,1,128,power_law_1.2,0.18176767349243164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,1,128,power_law_1.2,0.20517248153686524
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,1,128,power_law_1.2,0.2802112007141113
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,1,128,power_law_1.2,0.11662976264953613
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,1,128,power_law_1.2,0.36533374786376954
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,1,128,power_law_1.2,0.11980671882629394
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,1,128,power_law_1.2,0.6602483367919921
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,1,128,power_law_1.2,0.044788479804992676
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,1,128,power_law_1.2,0.8225421142578124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,1,128,power_law_1.2,0.04482560157775879
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,1,128,power_law_1.2,0.04468224048614502
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,1,128,power_law_1.2,0.059671039581298824
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,1,128,power_law_1.2,0.05960192203521728
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,1,128,power_law_1.2,0.0603110408782959
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,1,128,power_law_1.2,1.2906278991699218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,1,128,power_law_1.2,0.06113152027130127
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,1,128,power_law_1.2,0.061900801658630364
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,1,128,power_law_1.2,0.06309887886047363
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,1,128,power_law_1.2,0.06510848045349121
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,1,128,power_law_1.2,0.06465151786804199
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,1,128,power_law_1.2,0.06770304203033448
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,1,128,power_law_1.2,0.06609663963317872
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,1,128,power_law_1.2,1.7754432678222656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,1,128,power_law_1.2,0.07253888130187988
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,1,128,power_law_1.2,0.07674367904663086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,1,128,power_law_1.2,0.07688960075378418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,1,128,power_law_1.2,0.0899788761138916
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,1,128,power_law_1.2,0.10154111862182617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,1,128,power_law_1.2,0.13925503730773925
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,1,128,power_law_1.2,0.16687103271484377
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,1,128,power_law_1.2,0.29030656814575195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,1,128,power_law_1.2,2.578871154785156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,1,128,power_law_1.2,0.39780479431152344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,2,64,balanced,0.04490880012512207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,2,64,balanced,0.0309555196762085
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,1,128,power_law_1.2,0.5656345748901368
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,2,64,balanced,0.039142398834228515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,1,128,power_law_1.2,0.07858176231384277
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,2,64,balanced,0.05649407863616943
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,2,64,balanced,0.06692863941192627
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,1,128,power_law_1.2,0.8459340667724609
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,2,64,balanced,0.1106816005706787
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,2,64,balanced,0.11220735549926757
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,2,64,balanced,0.11339520454406737
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,2,64,balanced,0.11186047554016114
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,2,64,balanced,0.11282048225402833
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,2,64,balanced,0.11594880104064942
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,2,64,balanced,0.1154700756072998
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,2,64,balanced,0.11433728218078613
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,2,64,balanced,0.12030464172363282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,2,64,balanced,0.1214310359954834
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,2,64,balanced,0.12763263702392577
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,2,64,balanced,0.12968192100524903
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,2,64,balanced,0.12728063583374022
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,2,64,balanced,0.13416192054748535
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,2,64,balanced,0.14133760452270508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,2,64,balanced,0.030214400291442872
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,2,64,balanced,0.16011903762817384
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,2,64,balanced,0.16909183502197264
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,1,128,power_law_1.2,4.486134948730468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,2,64,balanced,0.20614271163940429
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,2,64,balanced,0.25059583663940427
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,2,64,balanced,0.02977407932281494
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,2,64,balanced,0.029583361148834232
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,2,64,balanced,0.030083839893341062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,2,64,balanced,0.3828083038330078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,2,64,balanced,0.033393919467926025
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,1,128,power_law_1.2,1.0772019195556641
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,2,64,balanced,0.4551001739501953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,2,64,balanced,0.05715456008911133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,2,64,balanced,0.05670527935028076
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,2,64,balanced,0.05777023792266846
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,2,64,balanced,0.05789951801300049
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,2,64,balanced,0.058428158760070806
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,2,64,balanced,0.06018303871154786
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,2,64,balanced,0.060756478309631344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,2,64,balanced,0.0631667184829712
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,2,64,balanced,0.06309375762939454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,2,64,balanced,0.06713856220245361
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,2,64,balanced,0.07388671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,2,64,balanced,0.06685056209564208
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,2,64,balanced,0.07397759914398193
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,2,64,balanced,0.08094207763671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,2,64,balanced,0.09038463592529297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,2,64,balanced,0.10177536010742187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,2,64,balanced,0.13049344062805174
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,2,64,balanced,0.15355648040771483
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,2,64,balanced,0.20253824234008788
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,2,64,balanced,0.02988287925720215
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,2,64,balanced,0.2528191947937012
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,2,64,power_law_1.01,0.066942720413208
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,2,64,balanced,0.044535040855407715
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,2,64,power_law_1.01,0.066494722366333
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,2,64,power_law_1.01,0.06673664093017578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,2,64,power_law_1.01,0.06715263843536376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,2,64,power_law_1.01,0.10639231681823731
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,2,64,power_law_1.01,0.1053171157836914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,2,64,power_law_1.01,0.11275903701782226
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,2,64,balanced,0.06429440021514893
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,2,64,power_law_1.01,0.11258624076843263
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,2,64,power_law_1.01,0.11281279563903808
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,2,64,power_law_1.01,0.11485695838928223
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,2,64,power_law_1.01,0.12053376197814943
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,2,64,power_law_1.01,0.12243328094482422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,2,64,power_law_1.01,0.12942975997924805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,2,64,power_law_1.01,0.1333529567718506
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,2,64,power_law_1.01,0.13011712074279785
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,2,64,power_law_1.01,0.13859583854675292
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,2,64,power_law_1.01,0.14951295852661134
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,2,64,power_law_1.01,0.16979072570800782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,2,64,power_law_1.01,0.21084672927856446
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,2,64,power_law_1.01,0.25414783477783204
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,2,64,power_law_1.01,0.3864857482910156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,2,64,power_law_1.01,0.11158783912658692
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,2,64,power_law_1.01,0.4426534271240234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,2,64,power_law_1.01,0.04361855983734131
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,2,64,power_law_1.01,0.6310630416870117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,2,64,power_law_1.01,0.043957757949829104
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,1,128,power_law_1.2,1.8698637390136716
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,2,64,power_law_1.01,0.04478720188140869
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,2,64,power_law_1.01,0.8538368225097657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,2,64,power_law_1.01,0.04432511806488037
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,2,64,power_law_1.01,0.055448322296142584
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,2,64,power_law_1.01,0.05935872077941895
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,2,64,power_law_1.01,0.056915202140808106
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,2,64,power_law_1.01,0.06108799934387207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,2,64,power_law_1.01,0.06177792072296142
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,2,64,power_law_1.01,1.2280537414550783
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,2,64,power_law_1.01,0.06284800052642822
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,2,64,power_law_1.01,0.06581120014190674
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,2,64,power_law_1.01,0.0686732816696167
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,2,64,power_law_1.01,0.07520512104034424
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,2,64,power_law_1.01,0.0751974391937256
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,2,64,power_law_1.01,0.07250175952911378
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,2,64,power_law_1.01,1.5735116577148438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,2,64,power_law_1.01,0.08253184318542481
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,2,64,power_law_1.01,0.09310336112976074
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,2,64,power_law_1.01,0.11645055770874024
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,2,64,power_law_1.01,0.13689599990844728
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,2,64,power_law_1.01,0.20241407394409178
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,2,64,power_law_1.01,0.2710963249206543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,2,64,power_law_1.01,0.3463692855834961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,2,64,power_law_1.2,0.0665228796005249
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,2,64,power_law_1.01,0.06389503955841065
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,2,64,power_law_1.01,0.4954291152954101
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,2,64,power_law_1.2,0.06604159832000732
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,2,64,power_law_1.2,0.06665855884552002
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,2,64,power_law_1.2,0.10454015731811524
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,2,64,power_law_1.01,0.7609664154052734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,2,64,power_law_1.2,0.10712191581726074
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,2,64,power_law_1.2,0.1109286403656006
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,2,64,power_law_1.2,0.11131775856018067
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,2,64,power_law_1.01,0.9019993591308595
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,2,64,power_law_1.2,0.11257216453552246
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,2,64,power_law_1.2,0.11943167686462401
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,2,64,power_law_1.2,0.12347135543823243
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,2,64,power_law_1.2,0.11293312072753907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,2,64,power_law_1.2,0.1273843193054199
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,2,64,power_law_1.2,0.12844672203063964
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,2,64,power_law_1.2,0.13405440330505372
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,2,64,power_law_1.01,0.05924223899841309
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,2,64,power_law_1.2,0.133818883895874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,2,64,power_law_1.2,0.14426624298095703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,2,64,power_law_1.2,0.16337152481079104
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,2,64,power_law_1.2,0.1788364791870117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,2,64,power_law_1.2,0.24732799530029298
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,2,64,power_law_1.2,0.3265305709838867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,2,64,power_law_1.2,0.43517696380615234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,2,64,power_law_1.2,0.10610688209533692
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,2,64,power_law_1.2,0.5690521621704102
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,2,64,power_law_1.2,0.044003839492797854
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,2,64,power_law_1.2,0.04393216133117676
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,2,64,power_law_1.2,0.05545472145080567
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,2,64,power_law_1.2,0.054480638504028324
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,2,64,power_law_1.2,0.9433023834228516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,2,64,power_law_1.2,0.05633152008056641
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,2,64,power_law_1.2,0.05951104164123535
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,2,64,power_law_1.2,0.05860352039337158
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,2,64,power_law_1.2,1.21193603515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,2,64,power_law_1.2,0.06100863933563232
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,2,64,power_law_1.2,0.06344319820404053
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,2,64,power_law_1.2,0.06524415969848632
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,2,64,power_law_1.2,0.06954880237579346
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,2,64,power_law_1.2,0.0723801612854004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,2,64,power_law_1.2,0.07002880096435546
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,2,64,power_law_1.2,0.07338624000549317
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,2,64,power_law_1.2,0.07752960205078126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,2,64,power_law_1.2,0.04364031791687012
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,2,64,power_law_1.2,0.08891776084899902
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,2,64,power_law_1.2,0.09928832054138184
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,2,64,power_law_1.2,1.9418572998046877
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,2,64,power_law_1.2,0.14036864280700684
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,2,64,power_law_1.2,0.18086015701293945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,2,64,power_law_1.2,0.23618688583374023
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,2,64,power_law_1.2,0.06061439990997315
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,2,64,power_law_1.2,0.4638630294799805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,4,32,balanced,0.02823040008544922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,2,64,power_law_1.2,2.5561920166015626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,4,32,balanced,0.028090879917144772
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,4,32,balanced,0.02864000082015991
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,2,64,power_law_1.2,0.6528371429443359
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,4,32,balanced,0.04732416152954101
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,4,32,balanced,0.05638400077819824
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,4,32,balanced,0.067325439453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,4,32,balanced,0.1093388843536377
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,4,32,balanced,0.10950528144836427
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,4,32,balanced,0.11520000457763672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,4,32,balanced,0.11145215988159181
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,2,64,power_law_1.2,1.092078094482422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,4,32,balanced,0.1159705638885498
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,4,32,balanced,0.11762559890747071
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,4,32,balanced,0.12026111602783203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,4,32,balanced,0.12393343925476075
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,2,64,power_law_1.2,1.3391436767578124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,2,64,power_law_1.2,0.3147583961486816
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,4,32,balanced,0.12830592155456544
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,4,32,balanced,0.13482111930847168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,4,32,balanced,0.14874367713928222
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,4,32,balanced,0.15763072013854978
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,4,32,balanced,0.1845734405517578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,4,32,balanced,0.20039295196533202
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,4,32,balanced,0.25218944549560546
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,4,32,balanced,0.32561790466308593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,4,32,balanced,0.028936960697174073
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,4,32,balanced,0.4883750534057617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,4,32,balanced,0.02857599973678589
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,4,32,balanced,0.11283200263977052
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,4,32,balanced,0.02858880043029785
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,4,32,balanced,0.5928985595703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,4,32,balanced,0.029024000167846682
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,4,32,balanced,0.03379839897155762
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,4,32,balanced,0.12906880378723146
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,4,32,balanced,0.047133440971374514
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,4,32,balanced,0.12120575904846191
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,4,32,balanced,0.06315648078918457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,4,32,balanced,0.062151679992675776
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,4,32,balanced,0.06417280197143554
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,4,32,balanced,0.06413568019866943
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,4,32,balanced,0.06500736236572266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,4,32,balanced,0.06684927940368653
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,4,32,balanced,0.06906112194061279
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,4,32,balanced,0.07209216117858887
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,4,32,balanced,0.07452799797058104
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,4,32,balanced,0.0672166395187378
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,4,32,balanced,0.07291903972625732
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,4,32,balanced,0.07796095848083497
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,4,32,balanced,0.08624256134033204
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,4,32,balanced,0.0983961582183838
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,4,32,balanced,0.11968000411987305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,4,32,balanced,0.13975168228149415
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,4,32,balanced,0.18612096786499024
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,4,32,balanced,0.2244927978515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,4,32,power_law_1.01,0.06629759788513183
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,4,32,balanced,0.3091340827941894
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,4,32,power_law_1.01,0.06679423809051513
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,4,32,power_law_1.01,0.06703743934631348
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,4,32,balanced,0.39297279357910153
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,4,32,power_law_1.01,0.06730624198913573
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,4,32,power_law_1.01,0.10653056144714354
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,4,32,power_law_1.01,0.10053631782531738
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,4,32,power_law_1.01,0.10781824111938476
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,4,32,power_law_1.01,0.1106764793395996
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,4,32,power_law_1.01,0.11407872200012208
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,4,32,power_law_1.01,0.11594752311706542
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,4,32,power_law_1.01,0.11862015724182129
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,4,32,balanced,0.0651366376876831
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,4,32,power_law_1.01,0.12698623657226563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,4,32,power_law_1.01,0.12508031845092774
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,4,32,power_law_1.01,0.1280396842956543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,4,32,power_law_1.01,0.1294502353668213
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,4,32,power_law_1.01,0.15031295776367187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,4,32,power_law_1.01,0.16253055572509764
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,4,32,power_law_1.01,0.19251968383789062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,4,32,power_law_1.01,0.23193727493286134
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,4,32,power_law_1.01,0.31825408935546873
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,4,32,power_law_1.01,0.37461761474609373
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,4,32,power_law_1.01,0.04620416164398193
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,4,32,power_law_1.01,0.046197757720947266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,4,32,power_law_1.01,0.10751615524291994
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,4,32,power_law_1.01,0.6918476867675781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,4,32,power_law_1.01,0.12177151679992675
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,4,32,power_law_1.01,0.04742015838623047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,4,32,power_law_1.01,0.046210560798645016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,4,32,power_law_1.01,0.05733759880065918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,4,32,power_law_1.01,0.06234240055084228
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,4,32,power_law_1.01,0.06354688167572022
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,4,32,power_law_1.01,0.06708992004394532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,4,32,power_law_1.01,0.07016960144042969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,4,32,power_law_1.01,0.07279488086700439
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,4,32,power_law_1.01,0.5237350463867188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,4,32,power_law_1.01,0.07422336101531982
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,4,32,power_law_1.01,0.07081088066101074
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,4,32,power_law_1.01,0.07270400047302246
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,4,32,power_law_1.01,0.046337280273437496
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,4,32,power_law_1.01,0.07683584213256836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,4,32,power_law_1.01,0.08639488220214844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,4,32,power_law_1.01,0.09901951789855958
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,4,32,power_law_1.01,1.0103539276123048
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,4,32,power_law_1.01,0.12706175804138184
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,4,32,power_law_1.01,0.06333568096160888
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,4,32,power_law_1.01,0.14941311836242677
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,4,32,power_law_1.01,0.06478464126586914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,4,32,power_law_1.01,0.20235904693603515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,4,32,power_law_1.01,0.2269977569580078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,4,32,power_law_1.01,1.4761727905273436
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,4,32,power_law_1.2,0.06638207912445068
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,4,32,power_law_1.01,0.37046142578125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,4,32,power_law_1.2,0.06654975891113282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,4,32,power_law_1.01,0.46689025878906254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,4,32,power_law_1.2,0.06781055927276611
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,4,32,power_law_1.2,0.10471679687499999
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,4,32,power_law_1.01,0.7083033752441407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,4,32,power_law_1.2,0.09676032066345215
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,4,32,power_law_1.2,0.10569472312927246
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,4,32,power_law_1.2,0.11265536308288575
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,4,32,power_law_1.2,0.10785280227661134
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,4,32,power_law_1.01,0.9291264343261719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,4,32,power_law_1.2,0.11492992401123048
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,4,32,power_law_1.2,0.12043392181396484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,4,32,power_law_1.2,0.12462592124938965
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,4,32,power_law_1.2,0.12662143707275392
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,4,32,power_law_1.2,0.13033472061157228
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,4,32,power_law_1.2,0.14018048286437987
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,4,32,power_law_1.2,0.1551027202606201
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,4,32,power_law_1.2,0.15816320419311522
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,4,32,power_law_1.2,0.1967910385131836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,4,32,power_law_1.2,0.06858240127563477
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,4,32,power_law_1.2,0.25396863937377934
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,4,32,power_law_1.2,0.36548480987548826
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,4,32,power_law_1.2,0.46244350433349607
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,4,32,power_law_1.2,0.11706239700317382
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,4,32,power_law_1.2,0.04611968040466309
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,4,32,power_law_1.2,0.7313024139404296
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,4,32,power_law_1.2,0.04652927875518799
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,4,32,power_law_1.2,0.046407680511474605
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,4,32,power_law_1.2,0.9140940856933593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,4,32,power_law_1.2,0.06044672012329102
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,4,32,power_law_1.2,0.06123775959014892
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,4,32,power_law_1.2,0.0636352014541626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,4,32,power_law_1.2,0.06388991832733154
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,4,32,power_law_1.2,1.4636250305175782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,4,32,power_law_1.2,0.045920000076293946
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,4,32,power_law_1.2,0.06744832038879395
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,4,32,power_law_1.2,0.07057407855987549
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,4,32,power_law_1.2,0.073503999710083
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,4,32,power_law_1.2,0.07242623805999757
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,4,32,power_law_1.2,0.0718886423110962
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,4,32,power_law_1.2,0.12794367790222166
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,4,32,power_law_1.2,1.9425318908691405
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,4,32,power_law_1.2,0.07852287769317627
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,4,32,power_law_1.2,0.09245311737060546
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,4,32,power_law_1.2,0.10063615798950196
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,4,32,power_law_1.2,0.12306943893432618
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,4,32,power_law_1.2,0.06222976207733154
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,4,32,power_law_1.2,0.15636608123779297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,4,32,power_law_1.2,0.23163904190063475
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,4,32,power_law_1.2,0.2951475143432617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,8,16,balanced,0.02799488067626953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,4,32,power_law_1.2,0.065829119682312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,4,32,power_law_1.2,0.4145151901245117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,8,16,balanced,0.02746495962142944
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,8,16,balanced,0.028887040615081787
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,8,16,balanced,0.04774784088134766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,8,16,balanced,0.05666944026947022
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,8,16,balanced,0.07390207767486572
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,4,32,power_law_1.2,0.6462169647216797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,8,16,balanced,0.11468928337097169
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,8,16,balanced,0.11467007637023925
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,8,16,balanced,0.11855999946594238
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,8,16,balanced,0.11842816352844239
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,8,16,balanced,0.12162048339843749
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,8,16,balanced,0.12390399932861329
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,8,16,balanced,0.1271731185913086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,4,32,power_law_1.2,0.9436595153808593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,8,16,balanced,0.1202892780303955
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,8,16,balanced,0.12395903587341309
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,8,16,balanced,0.12698623657226563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,8,16,balanced,0.13183744430541994
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,8,16,balanced,0.1422489643096924
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,8,16,balanced,0.171015682220459
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,8,16,balanced,0.18921472549438476
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,8,16,balanced,0.23210111618041993
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,4,32,power_law_1.2,1.2562380981445311
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,8,16,balanced,0.2685030364990234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,8,16,balanced,0.3543052673339844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,8,16,balanced,0.02842751979827881
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,8,16,balanced,0.4559936141967773
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,8,16,balanced,0.0288319993019104
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,8,16,balanced,0.030380799770355228
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,8,16,balanced,0.038460159301757814
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,8,16,balanced,0.6941426849365234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,8,16,balanced,0.07083903789520264
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,8,16,balanced,0.07071231842041016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,8,16,balanced,0.07292287826538087
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,8,16,balanced,0.07241087913513183
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,8,16,balanced,0.872933120727539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,8,16,balanced,0.0753984022140503
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,8,16,balanced,0.14991488456726074
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,8,16,balanced,0.07827328205108643
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,8,16,balanced,0.08050432205200195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,8,16,balanced,0.07289472103118896
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,8,16,balanced,0.0761356782913208
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,8,16,balanced,0.07815552234649659
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,8,16,balanced,0.08231552124023438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,8,16,balanced,0.09021568298339844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,8,16,balanced,0.10081664085388184
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,8,16,balanced,0.028400640487670898
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,8,16,balanced,0.11973888397216796
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,8,16,balanced,0.1395315170288086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,4,32,power_law_1.2,0.07520127773284913
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,8,16,balanced,0.1760537528991699
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,8,16,balanced,0.21227647781372072
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,8,16,balanced,0.05089663982391357
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,8,16,balanced,0.29834880828857424
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,8,16,power_law_1.01,0.05555712223052979
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,8,16,balanced,0.36930049896240236
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,8,16,power_law_1.01,0.06761856079101562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,8,16,power_law_1.01,0.06609792232513427
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,8,16,power_law_1.01,0.06885759830474854
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,8,16,balanced,0.5262502288818359
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,8,16,power_law_1.01,0.06968959808349609
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,8,16,power_law_1.01,0.10565119743347169
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,8,16,balanced,0.6863692474365235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,8,16,power_law_1.01,0.1106764793395996
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,8,16,power_law_1.01,0.11629823684692382
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,8,16,power_law_1.01,0.12240639686584473
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,8,16,power_law_1.01,0.12468352317810058
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,8,16,power_law_1.01,0.12440832138061524
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,8,16,power_law_1.01,0.12790016174316404
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,8,16,power_law_1.01,0.1283340835571289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,8,16,power_law_1.01,0.13105024337768553
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,8,16,power_law_1.01,0.13507072448730467
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,8,16,power_law_1.01,0.15138431549072268
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,8,16,power_law_1.01,0.16778240203857422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,8,16,power_law_1.01,0.20134912490844728
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,8,16,power_law_1.01,0.21411840438842775
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,8,16,power_law_1.01,0.29406335830688474
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,8,16,power_law_1.01,0.390332145690918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,8,16,power_law_1.01,0.11600000381469726
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,8,16,power_law_1.01,0.49991935729980475
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,8,16,power_law_1.01,0.04829567909240723
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,8,16,power_law_1.01,0.04833024024963379
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,8,16,power_law_1.01,0.6999565124511719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,8,16,power_law_1.01,0.04872704029083252
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,8,16,power_law_1.01,0.05062784194946289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,8,16,power_law_1.01,0.06474239826202392
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,8,16,power_law_1.01,0.9951667022705077
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,8,16,power_law_1.01,0.0679859209060669
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,8,16,power_law_1.01,0.06839168071746826
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,8,16,power_law_1.01,0.07066880226135254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,8,16,power_law_1.01,0.07835008144378662
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,8,16,power_law_1.01,0.07610112190246582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,8,16,power_law_1.01,1.200299530029297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,8,16,power_law_1.01,0.11125247955322266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,8,16,power_law_1.01,0.08006400108337403
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,8,16,power_law_1.01,0.0751692819595337
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,8,16,power_law_1.01,0.07881855964660645
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,8,16,power_law_1.01,0.10195839881896973
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,8,16,power_law_1.01,0.11621760368347167
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,8,16,power_law_1.01,0.14352767944335937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,8,16,power_law_1.01,0.16956031799316407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,8,16,power_law_1.01,0.21992704391479495
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,8,16,power_law_1.01,0.06920447826385498
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,8,16,power_law_1.01,0.2737766456604004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,8,16,power_law_1.01,0.4078041458129883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,8,16,power_law_1.2,0.05568895816802979
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,8,16,power_law_1.01,0.548331527709961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,8,16,power_law_1.01,0.036744959354400634
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,8,16,power_law_1.01,0.08150272369384766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,8,16,power_law_1.2,0.06388351917266846
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,8,16,power_law_1.01,0.08395263671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,8,16,power_law_1.2,0.06750080108642578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,8,16,power_law_1.01,0.75388671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,8,16,power_law_1.2,0.07006847858428955
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,8,16,power_law_1.2,0.10350848197937013
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,8,16,power_law_1.2,0.11129983901977539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,8,16,power_law_1.01,0.9012544250488281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,8,16,power_law_1.2,0.11584128379821776
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,8,16,power_law_1.2,0.12346367835998535
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,8,16,power_law_1.2,0.11786751747131348
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,8,16,power_law_1.2,0.12806528091430663
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,8,16,power_law_1.2,0.1294054412841797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,8,16,power_law_1.2,0.12852479934692382
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,8,16,power_law_1.2,0.12270336151123047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,8,16,power_law_1.2,0.1296115207672119
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,8,16,power_law_1.2,0.1386240005493164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,8,16,power_law_1.2,0.15066880226135254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,8,16,power_law_1.2,0.17332096099853517
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,8,16,power_law_1.2,0.20817792892456052
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,8,16,power_law_1.2,0.06746623992919923
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,8,16,power_law_1.2,0.24944896697998048
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,8,16,power_law_1.2,0.3650265502929687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,8,16,power_law_1.2,0.4413324737548828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,8,16,power_law_1.2,0.036524798870086674
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,8,16,power_law_1.2,0.10992256164550782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,8,16,power_law_1.2,0.5903180694580078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,8,16,power_law_1.2,0.047813119888305666
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,8,16,power_law_1.2,0.04636415958404541
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,8,16,power_law_1.2,0.047019519805908204
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,8,16,power_law_1.2,0.8681126403808594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,8,16,power_law_1.2,0.04990975856781006
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,8,16,power_law_1.2,0.06391551971435547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,8,16,power_law_1.2,0.06844031810760498
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,8,16,power_law_1.2,0.06842112064361572
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,8,16,power_law_1.2,0.07264895915985108
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,8,16,power_law_1.2,0.0765222406387329
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,8,16,power_law_1.2,0.07880576133728028
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,8,16,power_law_1.2,0.08077823638916015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,8,16,power_law_1.2,0.07544832229614258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,8,16,power_law_1.2,0.07965824127197266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,8,16,power_law_1.2,0.0807372760772705
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,8,16,power_law_1.2,0.08677120208740234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,8,16,power_law_1.2,1.5518028259277343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,8,16,power_law_1.2,0.1013145637512207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,8,16,power_law_1.2,0.11615232467651367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,8,16,power_law_1.2,0.15180928230285645
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,8,16,power_law_1.2,0.17677312850952148
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,8,16,power_law_1.2,0.2686220741271973
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,8,16,power_law_1.2,0.31361663818359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,8,16,power_law_1.2,0.06991871833801269
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,8,16,power_law_1.2,1.115898895263672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,8,16,power_law_1.2,0.4601855850219726
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,16,8,balanced,0.027933440208435058
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,16,8,balanced,0.029221119880676272
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,16,8,balanced,0.0331443190574646
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,16,8,balanced,0.061205759048461914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,16,8,balanced,0.05144063949584961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,16,8,balanced,0.08340736389160155
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,16,8,balanced,0.11775232315063477
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,16,8,balanced,0.12198399543762206
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,16,8,balanced,0.12472576141357421
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,16,8,balanced,0.1279193592071533
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,8,16,power_law_1.2,0.914919662475586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,16,8,balanced,0.12393343925476075
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,16,8,balanced,0.12626943588256836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,16,8,balanced,0.13081600189208983
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,16,8,balanced,0.1358463954925537
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,16,8,balanced,0.14972288131713868
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,16,8,balanced,0.16746240615844726
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,16,8,balanced,0.18471935272216797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,8,16,power_law_1.2,1.3144166564941406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,16,8,balanced,0.22074111938476562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,16,8,balanced,0.25723520278930667
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,16,8,balanced,0.3372313690185547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,16,8,balanced,0.40807937622070317
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,16,8,balanced,0.11995136260986328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,16,8,balanced,0.02867840051651001
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,16,8,balanced,0.5608000183105469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,16,8,balanced,0.7358386993408204
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,16,8,balanced,0.029201920032501223
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,16,8,balanced,0.1425216007232666
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,16,8,balanced,0.03464576005935669
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,16,8,balanced,0.04503680229187011
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,16,8,balanced,0.08539520263671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,16,8,balanced,0.08498175621032715
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,16,8,balanced,1.131156463623047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,16,8,balanced,0.08980352401733398
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,16,8,balanced,0.08431872367858886
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,8,16,power_law_1.2,0.6320665740966798
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,16,8,balanced,0.087325439453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,16,8,balanced,0.08992768287658691
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,16,8,balanced,0.09346431732177735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,16,8,balanced,0.10090880393981934
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,16,8,balanced,0.09661952018737793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,16,8,balanced,1.4608691406249998
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,16,8,balanced,0.12803199768066406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,16,8,balanced,0.14615424156188966
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,16,8,balanced,0.1821913528442383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,16,8,balanced,0.03134848117828369
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,16,8,balanced,0.2186227226257324
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,16,8,balanced,0.2888832092285156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,16,8,balanced,0.06428927898406982
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,16,8,balanced,0.35696510314941404
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,16,8,balanced,0.09300352096557617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,16,8,power_law_1.01,0.044546561241149904
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,16,8,balanced,0.5225971221923829
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,16,8,power_law_1.01,0.05764224052429199
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,16,8,balanced,0.6596300506591797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,16,8,power_law_1.01,0.05986048221588135
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,16,8,power_law_1.01,0.0633676815032959
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,16,8,balanced,0.11077631950378417
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,16,8,power_law_1.01,0.06962560176849365
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,16,8,power_law_1.01,0.06981120109558106
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,16,8,power_law_1.01,0.11484160423278808
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,16,8,power_law_1.01,0.10924160003662109
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,16,8,power_law_1.01,0.11840383529663086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,16,8,power_law_1.01,0.12602751731872558
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,16,8,power_law_1.01,0.12425472259521483
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,16,8,power_law_1.01,0.12079104423522949
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,16,8,power_law_1.01,0.12867072105407715
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,16,8,balanced,1.2700723266601561
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,16,8,power_law_1.01,0.13255167961120606
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,16,8,power_law_1.01,0.13913344383239745
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,16,8,power_law_1.01,0.15593983650207519
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,16,8,power_law_1.01,0.1738163185119629
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,16,8,power_law_1.01,0.204769287109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,16,8,power_law_1.01,0.2437900733947754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,16,8,power_law_1.01,0.335338249206543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,16,8,power_law_1.01,0.3761203384399414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,16,8,balanced,0.9643468475341797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,16,8,power_law_1.01,0.12102016448974609
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,16,8,power_law_1.01,0.5870809555053711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,16,8,power_law_1.01,0.03321727991104126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,16,8,power_law_1.01,0.04018688201904297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,16,8,power_law_1.01,0.7887398529052734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,16,8,power_law_1.01,0.04342016220092774
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,16,8,power_law_1.01,0.055957760810852054
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,16,8,power_law_1.01,0.05607295989990234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,16,8,power_law_1.01,0.07741439819335938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,16,8,power_law_1.01,1.0962547302246093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,16,8,power_law_1.01,0.07932672023773193
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,16,8,power_law_1.01,0.08331392288208009
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,16,8,power_law_1.01,0.08483200073242188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,16,8,power_law_1.01,0.08756863594055175
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,16,8,power_law_1.01,1.3747506713867188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,16,8,power_law_1.01,0.12597503662109374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,16,8,power_law_1.01,0.08862719535827637
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,16,8,power_law_1.01,0.08891136169433593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,16,8,power_law_1.01,0.09067263603210449
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,16,8,power_law_1.01,0.09784959793090821
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,16,8,power_law_1.01,0.09306624412536621
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,16,8,power_law_1.01,0.05215487957000733
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,16,8,power_law_1.01,0.1385548782348633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,16,8,power_law_1.01,0.16706176757812502
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,16,8,power_law_1.01,0.19962240219116212
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,16,8,power_law_1.01,0.283448314666748
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,16,8,power_law_1.01,0.3269388961791992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,16,8,power_law_1.01,0.4671744155883789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,16,8,power_law_1.01,0.09094655990600586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,16,8,power_law_1.2,0.03808128118515015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,16,8,power_law_1.2,0.05742208003997803
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,16,8,power_law_1.01,0.5896319961547851
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,16,8,power_law_1.2,0.05563648223876953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,16,8,power_law_1.2,0.06939008235931396
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,16,8,power_law_1.2,0.06676864147186279
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,16,8,power_law_1.01,0.11967103958129884
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,16,8,power_law_1.2,0.10835840225219726
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,16,8,power_law_1.2,0.11436544418334962
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,16,8,power_law_1.2,0.11683199882507325
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,16,8,power_law_1.2,0.12140800476074218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,16,8,power_law_1.01,0.9236505889892579
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,16,8,power_law_1.2,0.12495871543884278
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,16,8,power_law_1.2,0.12114175796508789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,16,8,power_law_1.2,0.12137855529785155
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,16,8,power_law_1.2,0.12595328330993652
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,16,8,power_law_1.01,1.0754611206054687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,16,8,power_law_1.2,0.13629695892333984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,16,8,power_law_1.2,0.14481023788452146
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,16,8,power_law_1.2,0.16211584091186523
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,16,8,power_law_1.2,0.17966463088989257
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,16,8,power_law_1.2,0.22197504043579103
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,16,8,power_law_1.2,0.25941120147705077
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,16,8,power_law_1.2,0.06213759899139404
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,16,8,power_law_1.2,0.3426393508911133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,16,8,power_law_1.2,0.43638526916503906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,16,8,power_law_1.2,0.03321216106414795
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,16,8,power_law_1.2,0.6081459045410156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,16,8,power_law_1.2,0.03984639883041382
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,16,8,power_law_1.2,0.040814080238342286
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,16,8,power_law_1.2,0.0478169584274292
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,16,8,power_law_1.2,0.13054719924926758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,16,8,power_law_1.2,0.8160499572753906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,16,8,power_law_1.2,0.05346432209014893
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,16,8,power_law_1.2,0.05368959903717041
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,16,8,power_law_1.2,0.0766758394241333
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,16,8,power_law_1.2,0.08033280372619629
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,16,8,power_law_1.2,0.08846336364746094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,16,8,power_law_1.2,0.08576767921447755
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,16,8,power_law_1.2,1.1755251312255859
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,16,8,power_law_1.2,0.08375295639038086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,16,8,power_law_1.2,0.08863103866577149
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,16,8,power_law_1.2,0.0918131160736084
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,16,8,power_law_1.2,0.09577343940734864
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,16,8,power_law_1.2,1.348865203857422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,16,8,power_law_1.2,0.12135807991027832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,16,8,power_law_1.2,0.10488832473754883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,16,8,power_law_1.2,0.1365337562561035
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,16,8,power_law_1.2,0.17407871246337892
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,16,8,power_law_1.2,0.22015743255615233
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,16,8,power_law_1.2,0.29549823760986327
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,16,8,power_law_1.2,0.36067455291748046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,32,4,balanced,0.027048959732055667
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,16,8,power_law_1.2,0.5374553680419922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,16,8,power_law_1.2,0.08544128417968751
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,32,4,balanced,0.02931839942932129
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,32,4,balanced,0.03372672080993652
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,16,8,power_law_1.2,0.6665049743652344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,32,4,balanced,0.056655359268188474
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,32,4,balanced,0.06556032180786134
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,32,4,balanced,0.08520575523376464
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,32,4,balanced,0.0854361629486084
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,32,4,balanced,0.12063103675842286
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,32,4,balanced,0.12264448165893556
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,32,4,balanced,0.12569727897644042
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,32,4,balanced,0.12756735801696778
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,16,8,power_law_1.2,0.9572441864013672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,32,4,balanced,0.13058943748474122
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,32,4,balanced,0.1357094383239746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,32,4,balanced,0.1382323169708252
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,32,4,balanced,0.14425984382629395
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,32,4,balanced,0.15094911575317382
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,32,4,balanced,0.1685273551940918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,32,4,balanced,0.18461952209472657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,32,4,balanced,0.22079103469848632
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,32,4,balanced,0.25746047973632813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,32,4,balanced,0.3386726379394531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,32,4,balanced,0.41156734466552736
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,32,4,balanced,0.0868671989440918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,32,4,balanced,0.5619353485107422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,32,4,balanced,0.028309760093688963
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,32,4,balanced,0.7305101013183594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,32,4,balanced,0.02942591905593872
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,32,4,balanced,0.03302783966064453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,16,8,power_law_1.2,1.3254899597167968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,32,4,balanced,0.041173758506774905
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,32,4,balanced,0.07936128139495849
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,16,8,power_law_1.2,0.08176511764526367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,32,4,balanced,0.08290687561035157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,32,4,balanced,1.1301427459716797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,32,4,balanced,0.10785535812377929
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,32,4,balanced,0.10921088218688965
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,32,4,balanced,0.11189120292663575
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,32,4,balanced,0.11377152442932129
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,32,4,balanced,0.11625727653503418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,32,4,balanced,0.12063360214233398
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,32,4,balanced,0.12147583961486816
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,32,4,balanced,0.12736384391784666
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,32,4,balanced,0.1354931163787842
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,32,4,balanced,1.5127693176269532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,32,4,balanced,0.1543449592590332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,32,4,balanced,0.17033472061157226
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,32,4,balanced,0.2047475242614746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,32,4,balanced,0.052925438880920404
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,32,4,balanced,0.23646976470947267
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,32,4,balanced,0.30509952545166014
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,32,4,balanced,0.37171585083007813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,32,4,balanced,0.0837491226196289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,32,4,power_law_1.01,0.030186240673065183
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,32,4,balanced,0.5557772827148437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,32,4,power_law_1.01,0.04922880172729492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,32,4,power_law_1.01,0.05495168209075928
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,32,4,power_law_1.01,0.0684876823425293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,32,4,balanced,0.6908415985107422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,32,4,power_law_1.01,0.07980544090270995
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,32,4,power_law_1.01,0.08052096366882325
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,32,4,power_law_1.01,0.07525631904602051
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,32,4,power_law_1.01,0.11627648353576661
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,32,4,power_law_1.01,0.11671808242797851
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,32,4,power_law_1.01,0.11902463912963868
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,32,4,power_law_1.01,0.12209664344787599
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,32,4,power_law_1.01,0.13223551750183105
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,32,4,power_law_1.01,0.13372159957885743
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,32,4,power_law_1.01,0.13811584472656252
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,32,4,power_law_1.01,0.1455014419555664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,32,4,balanced,1.3624409484863282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,32,4,power_law_1.01,0.16645120620727538
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,32,4,power_law_1.01,0.18638463973999025
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,32,4,power_law_1.01,0.22211072921752928
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,32,4,power_law_1.01,0.06254079818725586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,32,4,power_law_1.01,0.25720191955566407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,32,4,power_law_1.01,0.3365081787109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,32,4,power_law_1.01,0.4070412826538086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,32,4,power_law_1.01,0.031498239040374756
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,32,4,balanced,1.0442470550537108
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,32,4,power_law_1.01,0.12451071739196777
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,32,4,power_law_1.01,0.5944319915771484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,32,4,power_law_1.01,0.03632512092590332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,32,4,power_law_1.01,0.043023362159729
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,32,4,power_law_1.01,0.7570956420898438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,32,4,power_law_1.01,0.061410560607910156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,32,4,power_law_1.01,0.07006976127624512
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,32,4,power_law_1.01,0.06554368019104004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,32,4,power_law_1.01,0.07315711975097657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,32,4,power_law_1.01,0.09881855964660644
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,32,4,power_law_1.01,1.1418841552734373
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,32,4,power_law_1.01,0.1058892822265625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,32,4,power_law_1.01,0.1162508773803711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,32,4,power_law_1.01,0.11619327545166017
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,32,4,power_law_1.01,1.5441944885253907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,32,4,power_law_1.01,0.12976384162902832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,32,4,power_law_1.01,0.15015168190002443
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,32,4,power_law_1.01,0.17051008224487305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,32,4,power_law_1.01,0.20937984466552734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,32,4,power_law_1.01,0.09864064216613769
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,32,4,power_law_1.01,0.10330240249633789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,32,4,power_law_1.01,0.32787456512451174
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,32,4,power_law_1.01,0.10904319763183594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,32,4,power_law_1.01,0.4239756774902344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,32,4,power_law_1.01,0.12050815582275391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,32,4,power_law_1.2,0.029687039852142334
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,32,4,power_law_1.01,0.6136320114135743
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,32,4,power_law_1.01,0.05156608104705811
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,32,4,power_law_1.2,0.05430784225463867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,32,4,power_law_1.01,0.7717414093017578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,32,4,power_law_1.2,0.06141568183898925
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,32,4,power_law_1.2,0.06653312206268311
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,32,4,power_law_1.01,0.25386112213134765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,32,4,power_law_1.2,0.07591551780700684
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,32,4,power_law_1.01,1.1528844451904297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,32,4,power_law_1.2,0.08078720092773437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,32,4,power_law_1.2,0.11361536026000976
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,32,4,power_law_1.2,0.11683584213256837
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,32,4,power_law_1.01,1.427105255126953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,32,4,power_law_1.2,0.11978112220764162
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,32,4,power_law_1.2,0.122225923538208
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,32,4,power_law_1.2,0.12547455787658693
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,32,4,power_law_1.2,0.13090815544128417
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,32,4,power_law_1.2,0.045438718795776364
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,32,4,power_law_1.2,0.13924863815307617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,32,4,power_law_1.2,0.13477888107299804
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,32,4,power_law_1.2,0.14822015762329102
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,32,4,power_law_1.2,0.1681612777709961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,32,4,power_law_1.2,0.1882713508605957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,32,4,power_law_1.2,0.22583423614501955
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,32,4,power_law_1.2,0.2634611129760742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,32,4,power_law_1.2,0.34418174743652347
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,32,4,power_law_1.2,0.07942527770996094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,32,4,power_law_1.2,0.4208217620849609
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,32,4,power_law_1.2,0.031560959815979006
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,32,4,power_law_1.2,0.6338764953613281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,32,4,power_law_1.2,0.03608959913253784
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,32,4,power_law_1.2,0.04134399890899658
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,32,4,power_law_1.2,0.7931353759765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,32,4,power_law_1.2,0.04893311977386475
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,32,4,power_law_1.2,0.05613823890686035
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,32,4,power_law_1.2,0.06029695987701415
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,32,4,power_law_1.2,0.09501567840576172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,32,4,power_law_1.2,0.09702912330627442
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,32,4,power_law_1.2,0.10212863922119139
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,32,4,power_law_1.2,1.1571724700927735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,32,4,power_law_1.2,0.10663935661315918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,32,4,power_law_1.2,0.1105356788635254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,32,4,power_law_1.2,0.11351552009582519
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,32,4,power_law_1.2,0.11444479942321777
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,32,4,power_law_1.2,0.12119168281555175
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,32,4,power_law_1.2,1.5584486389160157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,32,4,power_law_1.2,0.13428223609924317
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,32,4,power_law_1.2,0.1527782440185547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,32,4,power_law_1.2,0.17432447433471682
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,32,4,power_law_1.2,0.21653760910034178
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,32,4,power_law_1.2,0.06704895973205567
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,32,4,power_law_1.2,0.2542835235595703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,32,4,power_law_1.2,0.07318784236907959
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,32,4,power_law_1.2,0.33036926269531247
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,1,128,balanced,0.017866239547729493
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,32,4,power_law_1.2,0.4327027130126953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,1,128,balanced,0.01780608057975769
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,1,128,balanced,0.017817599773406984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,1,128,balanced,0.017665280103683473
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,1,128,balanced,0.024151039123535157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,1,128,balanced,0.024311039447784424
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,1,128,balanced,0.024375040531158448
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,32,4,power_law_1.2,0.6235071945190429
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,1,128,balanced,0.02490623950958252
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,1,128,balanced,0.025011200904846192
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,32,4,power_law_1.2,0.787188491821289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,1,128,balanced,0.025487360954284666
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,1,128,balanced,0.02613503932952881
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,1,128,balanced,0.026548480987548827
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,1,128,balanced,0.02714751958847046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,1,128,balanced,0.02998784065246582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,1,128,balanced,0.030649600028991697
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,1,128,balanced,0.031788799762725826
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,1,128,balanced,0.03506047964096069
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,1,128,balanced,0.030533120632171628
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,1,128,balanced,0.03380223989486694
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,1,128,balanced,0.035361280441284185
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,1,128,balanced,0.04764544010162354
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,1,128,balanced,0.05653247833251953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,1,128,balanced,0.07532032012939453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,1,128,balanced,0.08752384185791015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,1,128,balanced,0.017923840284347535
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,1,128,balanced,0.02067199945449829
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,1,128,balanced,0.020706560611724854
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,1,128,balanced,0.020609281063079833
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,1,128,balanced,0.020471038818359374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,1,128,balanced,0.020719358921051027
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,1,128,balanced,0.02499840021133423
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,1,128,balanced,0.030379519462585446
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,1,128,balanced,0.026238720417022705
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,1,128,balanced,0.03067008018493652
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,1,128,balanced,0.030432000160217288
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,1,128,balanced,0.03118335962295532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,1,128,balanced,0.030979840755462645
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,1,128,balanced,0.031069440841674806
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,1,128,balanced,0.031870720386505125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,1,128,balanced,0.03146368026733398
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,1,128,balanced,0.03178240060806274
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,1,128,balanced,0.032231678962707525
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,32,4,power_law_1.2,1.1803826904296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,1,128,balanced,0.03249151945114136
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,1,128,balanced,0.03342848062515259
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,1,128,balanced,0.03438591957092285
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,1,128,balanced,0.0361075210571289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,1,128,balanced,0.03878655910491943
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,1,128,balanced,0.03603584051132202
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,1,128,balanced,0.03766527891159058
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,1,128,balanced,0.0474073600769043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,1,128,balanced,0.05267839908599854
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,1,128,balanced,0.06297344207763672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,32,4,power_law_1.2,1.4939736938476562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,1,128,power_law_1.01,0.01759232044219971
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,1,128,power_law_1.01,0.01802240014076233
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,1,128,power_law_1.01,0.024067840576171874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,1,128,power_law_1.01,0.02491136074066162
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,1,128,power_law_1.01,0.024915199279785156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,1,128,power_law_1.01,0.025484800338745117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,1,128,power_law_1.01,0.026155519485473632
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,1,128,power_law_1.01,0.028175361156463623
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,1,128,power_law_1.01,0.02923775911331177
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,1,128,power_law_1.01,0.0303436803817749
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,1,128,power_law_1.01,0.03151360034942627
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,1,128,power_law_1.01,0.034406399726867674
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,1,128,power_law_1.01,0.03521663904190063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,1,128,power_law_1.01,0.034781439304351805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,1,128,power_law_1.01,0.03439487934112549
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,1,128,balanced,0.040087041854858396
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,1,128,power_law_1.01,0.043425278663635256
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,1,128,balanced,0.07495039939880371
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,1,128,power_law_1.01,0.05752575874328614
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,1,128,power_law_1.01,0.06510208129882813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,1,128,power_law_1.01,0.0812172794342041
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,1,128,power_law_1.01,0.0179967999458313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,1,128,power_law_1.01,0.12104448318481445
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,1,128,power_law_1.01,0.14063360214233397
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,1,128,power_law_1.01,0.22811264038085938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,1,128,power_law_1.01,0.020582399368286132
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,1,128,power_law_1.01,0.28355583190917966
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,1,128,power_law_1.01,0.02071295976638794
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,1,128,power_law_1.01,0.03045248031616211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,1,128,power_law_1.01,0.03036288022994995
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,1,128,power_law_1.01,0.03096832036972046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,1,128,power_law_1.01,0.03110912084579468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,1,128,power_law_1.01,0.6402329254150391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,1,128,power_law_1.01,0.03167360067367554
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,1,128,power_law_1.01,0.03178112030029297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,1,128,power_law_1.01,0.032250878810882566
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,1,128,power_law_1.01,0.032688639163970946
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,1,128,power_law_1.01,0.033400321006774904
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,1,128,power_law_1.01,0.03508991956710815
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,1,128,power_law_1.01,0.037245440483093264
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,1,128,power_law_1.01,0.03958143949508667
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,1,128,power_law_1.01,0.040527358055114746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,1,128,power_law_1.01,0.037194240093231204
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,1,128,power_law_1.01,0.04179327964782715
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,1,128,power_law_1.01,0.04779647827148438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,1,128,power_law_1.01,0.05717504024505615
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,1,128,power_law_1.01,0.09529472351074218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,1,128,power_law_1.01,0.06655871868133545
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,1,128,power_law_1.01,0.10995072364807128
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,1,128,power_law_1.01,0.17627519607543946
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,1,128,power_law_1.01,0.5000281524658203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,1,128,power_law_1.01,0.25129600524902346
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,1,128,power_law_1.2,0.017768959999084472
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,1,128,power_law_1.2,0.017812479734420777
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,1,128,power_law_1.2,0.024167680740356447
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,1,128,power_law_1.01,0.3486783981323242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,1,128,power_law_1.2,0.024445440769195557
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,1,128,power_law_1.2,0.024975359439849854
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,1,128,power_law_1.01,0.03442431926727295
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,1,128,power_law_1.01,0.45566593170166014
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,1,128,power_law_1.2,0.025167360305786136
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,1,128,power_law_1.2,0.025987839698791503
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,1,128,power_law_1.2,0.026234879493713382
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,1,128,power_law_1.2,0.03024127960205078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,1,128,power_law_1.2,0.0311846399307251
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,1,128,power_law_1.2,0.03551232099533081
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,1,128,power_law_1.2,0.03358079910278321
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,1,128,power_law_1.2,0.03747584104537964
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,1,128,power_law_1.2,0.041900801658630374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,1,128,power_law_1.2,0.05831039905548095
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,1,128,power_law_1.01,0.02907007932662964
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,1,128,power_law_1.2,0.06759552001953126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,1,128,power_law_1.2,0.09354623794555664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,1,128,power_law_1.2,0.11753472328186035
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,1,128,power_law_1.2,0.20525312423706055
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,1,128,power_law_1.2,0.26215423583984376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,1,128,power_law_1.2,0.030199038982391357
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,1,128,power_law_1.2,0.029870080947875976
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,1,128,power_law_1.2,0.03431040048599243
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,1,128,power_law_1.2,0.020583679676055906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,1,128,power_law_1.2,0.03057663917541504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,1,128,power_law_1.2,0.03073024034500122
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,1,128,power_law_1.2,0.0312665605545044
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,1,128,power_law_1.2,0.03108992099761963
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,1,128,power_law_1.2,0.03162879943847656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,1,128,power_law_1.2,0.0329036808013916
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,1,128,power_law_1.2,0.03260799884796143
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,1,128,power_law_1.2,0.033441278934478756
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,1,128,power_law_1.2,0.033962240219116216
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,1,128,power_law_1.2,0.03563776016235352
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,1,128,power_law_1.2,0.3894579315185547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,1,128,power_law_1.2,0.03691263914108277
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,1,128,power_law_1.2,0.020532479286193846
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,1,128,power_law_1.2,0.039813120365142826
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,1,128,power_law_1.2,0.038021121025085446
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,1,128,power_law_1.2,0.0398528003692627
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,1,128,power_law_1.2,0.048739838600158694
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,1,128,power_law_1.2,0.058903040885925295
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,1,128,power_law_1.2,0.07101439952850341
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,1,128,power_law_1.2,0.0980684757232666
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,1,128,power_law_1.2,0.14950655937194823
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,1,128,power_law_1.2,0.20857471466064453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,1,128,power_law_1.2,0.03764607906341553
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,2,64,balanced,0.016605440378189087
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,2,64,balanced,0.01643519997596741
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,1,128,power_law_1.2,0.30046079635620115
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,2,64,balanced,0.01639296054840088
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,2,64,balanced,0.016249599456787108
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,2,64,balanced,0.016586240530014038
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,2,64,balanced,0.02236799955368042
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,2,64,balanced,0.022775039672851563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,2,64,balanced,0.022979838848114015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,2,64,balanced,0.02373120069503784
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,2,64,balanced,0.024300799369812012
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,2,64,balanced,0.024337921142578125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,2,64,balanced,0.02573440074920654
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,2,64,balanced,0.02506623983383179
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,2,64,balanced,0.02764672040939331
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,2,64,balanced,0.026123518943786624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,2,64,balanced,0.02700160026550293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,2,64,balanced,0.028782079219818114
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,2,64,balanced,0.03226752042770385
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,2,64,balanced,0.027920639514923094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,2,64,balanced,0.03136384010314942
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,2,64,balanced,0.03262207984924316
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,2,64,balanced,0.0420198392868042
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,2,64,balanced,0.050636801719665524
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,2,64,balanced,0.06462975978851318
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,2,64,balanced,0.0732096004486084
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,1,128,power_law_1.2,0.5345727920532226
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,2,64,balanced,0.09851136207580566
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,2,64,balanced,0.11831168174743652
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,2,64,balanced,0.019646719694137574
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,2,64,balanced,0.019279359579086303
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,2,64,balanced,0.019377919435501097
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,2,64,balanced,0.019870719909667968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,2,64,balanced,0.028728320598602298
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,2,64,balanced,0.029253120422363284
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,2,64,balanced,0.029061119556427
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,2,64,balanced,0.029744639396667483
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,2,64,balanced,0.029358079433441164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,2,64,balanced,0.030164480209350586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,2,64,balanced,0.030140159130096433
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,2,64,balanced,0.031027200222015383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,2,64,balanced,0.031239678859710695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,2,64,balanced,0.032025599479675294
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,2,64,balanced,0.01966848015785217
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,2,64,balanced,0.033559041023254396
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,2,64,balanced,0.03603327989578247
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,2,64,balanced,0.032442879676818845
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,1,128,power_law_1.2,0.4112435150146484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,2,64,balanced,0.03458048105239868
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,2,64,balanced,0.03694720029830932
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,2,64,balanced,0.042031359672546384
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,2,64,balanced,0.04733439922332764
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,2,64,balanced,0.05673088073730469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,2,64,balanced,0.06699520111083984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,2,64,balanced,0.08886655807495117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,1,128,power_law_1.2,0.8233740997314453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,2,64,balanced,0.11145471572875976
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,2,64,balanced,0.0314687991142273
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,2,64,power_law_1.01,0.016830719709396362
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,2,64,power_law_1.01,0.016812800168991088
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,2,64,power_law_1.01,0.022366719245910643
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,2,64,power_law_1.01,0.02289535999298096
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,2,64,power_law_1.01,0.024070401191711426
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,2,64,power_law_1.01,0.02418816089630127
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,2,64,power_law_1.01,0.02490623950958252
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,2,64,power_law_1.01,0.02572160005569458
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,2,64,power_law_1.01,0.027032320499420164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,2,64,power_law_1.01,0.02678783893585205
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,2,64,power_law_1.01,0.02790015935897827
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,2,64,power_law_1.01,0.028637440204620362
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,2,64,power_law_1.01,0.03138432025909424
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,2,64,power_law_1.01,0.03211519956588745
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,2,64,power_law_1.01,0.03047039985656738
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,2,64,power_law_1.01,0.03215872049331665
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,2,64,power_law_1.01,0.04074111938476562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,2,64,power_law_1.01,0.04986112117767334
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,2,64,power_law_1.01,0.061178879737854006
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,2,64,power_law_1.01,0.07707647800445556
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,2,64,power_law_1.01,0.016665600538253784
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,2,64,power_law_1.01,0.10028032302856445
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,2,64,power_law_1.01,0.11776384353637695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,2,64,power_law_1.01,0.17595008850097654
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,2,64,power_law_1.01,0.2317849540710449
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,2,64,power_law_1.01,0.019664640426635745
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,2,64,power_law_1.01,0.3879731369018555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,2,64,power_law_1.01,0.019875839948654175
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,2,64,power_law_1.01,0.028637440204620362
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,2,64,power_law_1.01,0.0289574408531189
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,2,64,power_law_1.01,0.029687039852142334
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,2,64,power_law_1.01,0.029648640155792237
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,2,64,power_law_1.01,0.030832641124725342
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,2,64,power_law_1.01,0.031220479011535646
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,2,64,power_law_1.01,0.031449599266052244
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,2,64,power_law_1.01,0.03298687934875488
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,1,128,power_law_1.2,0.6048064041137695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,2,64,power_law_1.01,0.03368576049804688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,2,64,power_law_1.01,0.019482879638671874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,2,64,power_law_1.01,0.0358515191078186
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,2,64,power_law_1.01,0.035143680572509765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,2,64,power_law_1.01,0.03518208026885986
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,2,64,power_law_1.01,0.036432640552520754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,2,64,power_law_1.01,0.5105011367797851
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,2,64,power_law_1.01,0.040550398826599124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,2,64,power_law_1.01,0.04560256004333496
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,2,64,power_law_1.01,0.05793536186218262
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,2,64,power_law_1.01,0.07039743900299072
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,2,64,power_law_1.01,0.10008576393127441
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,2,64,power_law_1.01,0.11811584472656249
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,2,64,power_law_1.01,0.17254655838012695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,2,64,power_law_1.01,0.22827520370483398
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,2,64,power_law_1.2,0.01647104024887085
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,1,128,power_law_1.2,0.7763929748535157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,2,64,power_law_1.2,0.016806399822235106
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,2,64,power_law_1.01,0.3586265563964844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,2,64,power_law_1.01,0.4251839828491211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,2,64,power_law_1.2,0.01695232033729553
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,2,64,power_law_1.2,0.023432960510253904
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,2,64,power_law_1.2,0.022691841125488284
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,2,64,power_law_1.01,0.03586303949356079
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,2,64,power_law_1.2,0.024738559722900393
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,2,64,power_law_1.2,0.02704639911651611
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,2,64,power_law_1.2,0.028936960697174073
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,2,64,power_law_1.2,0.029533441066741946
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,2,64,power_law_1.2,0.03164671897888184
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,2,64,power_law_1.2,0.03302016019821167
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,2,64,power_law_1.2,0.030338559150695797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,2,64,power_law_1.2,0.03229439973831177
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,2,64,power_law_1.2,0.042817277908325194
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,2,64,power_law_1.2,0.0537011194229126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,2,64,power_law_1.2,0.06177408218383788
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,2,64,power_law_1.2,0.08154496192932129
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,2,64,power_law_1.2,0.09336959838867187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,2,64,power_law_1.2,0.15771136283874512
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,2,64,power_law_1.2,0.20385023117065432
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,2,64,power_law_1.2,0.023880960941314696
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,2,64,power_law_1.2,0.025143039226531983
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,2,64,power_law_1.2,0.026152958869934084
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,2,64,power_law_1.2,0.019595520496368406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,2,64,power_law_1.2,0.01982208013534546
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,2,64,power_law_1.2,0.01949056029319763
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,2,64,power_law_1.2,0.029063680171966554
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,2,64,power_law_1.2,0.029268479347229003
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,2,64,power_law_1.2,0.029662721157073975
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,2,64,power_law_1.2,0.030443520545959474
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,2,64,power_law_1.2,0.03113215923309326
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,2,64,power_law_1.2,0.031381759643554684
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,2,64,power_law_1.2,0.03203455924987793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,2,64,power_law_1.2,0.03373183965682983
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,2,64,power_law_1.2,0.03382911920547486
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,2,64,power_law_1.2,0.03515647888183594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,2,64,power_law_1.2,0.034603519439697264
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,2,64,power_law_1.2,0.29915775299072267
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,2,64,power_law_1.2,0.03670399904251099
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,2,64,power_law_1.2,0.03979903936386108
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,2,64,power_law_1.2,0.04708864212036133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,2,64,power_law_1.2,0.054434561729431154
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,2,64,power_law_1.2,0.07123199939727784
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,2,64,power_law_1.2,0.08905599594116212
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,2,64,power_law_1.2,0.12963839530944826
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,2,64,power_law_1.2,0.16471040725708008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,2,64,power_law_1.2,0.0367411208152771
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,2,64,power_law_1.2,0.24210559844970705
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,4,32,balanced,0.016020480394363403
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,4,32,balanced,0.016144640445709228
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,4,32,balanced,0.015813119411468506
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,4,32,balanced,0.016469759941101076
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,4,32,balanced,0.017059839963912964
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,4,32,balanced,0.022671360969543457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,4,32,balanced,0.023567359447479248
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,4,32,balanced,0.0240883207321167
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,4,32,balanced,0.024883201122283934
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,4,32,balanced,0.025429759025573727
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,4,32,balanced,0.02671744108200073
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,4,32,balanced,0.02566015958786011
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,4,32,balanced,0.025987839698791503
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,4,32,balanced,0.02799488067626953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,4,32,balanced,0.029255681037902832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,2,64,power_law_1.2,0.40532993316650395
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,4,32,balanced,0.03036799907684326
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,4,32,balanced,0.026740479469299316
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,4,32,balanced,0.02923392057418823
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,4,32,balanced,0.03135231971740723
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,4,32,balanced,0.04085887908935547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,4,32,balanced,0.048565759658813476
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,4,32,balanced,0.05963776111602783
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,4,32,balanced,0.06948480129241943
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,4,32,balanced,0.08924287796020508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,4,32,balanced,0.10658047676086427
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,4,32,balanced,0.14772992134094237
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,4,32,balanced,0.1867647933959961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,4,32,balanced,0.019426560401916503
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,4,32,balanced,0.019340800046920775
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,4,32,balanced,0.01930367946624756
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,4,32,balanced,0.019842560291290286
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,4,32,balanced,0.02014847993850708
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,4,32,balanced,0.029283840656280518
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,4,32,balanced,0.030013439655303953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,4,32,balanced,0.03053056001663208
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,4,32,balanced,0.03125119924545288
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,4,32,balanced,0.031732480525970455
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,4,32,balanced,0.03287168025970459
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,2,64,power_law_1.2,0.6141823959350586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,4,32,balanced,0.03127808094024658
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,2,64,power_law_1.2,0.3284979248046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,4,32,balanced,0.03191807985305786
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,4,32,balanced,0.0332915210723877
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,4,32,balanced,0.034275839328765864
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,4,32,balanced,0.035539200305938726
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,4,32,balanced,0.03157119989395142
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,4,32,balanced,0.03355776071548462
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,4,32,balanced,0.03699712038040161
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,4,32,balanced,0.04062079906463623
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,4,32,balanced,0.0444979190826416
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,4,32,balanced,0.054456319808959965
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,4,32,balanced,0.0641919994354248
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,4,32,balanced,0.08451071739196778
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,1,128,power_law_1.2,1.106682891845703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,4,32,balanced,0.10472703933715821
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,4,32,balanced,0.14278143882751465
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,4,32,balanced,0.17860095977783202
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,4,32,power_law_1.01,0.016437760591506957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,4,32,power_law_1.01,0.016743680238723753
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,4,32,power_law_1.01,0.01705855965614319
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,4,32,power_law_1.01,0.022396159172058106
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,4,32,power_law_1.01,0.023584001064300537
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,4,32,power_law_1.01,0.025414400100708008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,4,32,power_law_1.01,0.025108480453491212
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,4,32,power_law_1.01,0.026708478927612307
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,4,32,power_law_1.01,0.027905280590057376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,4,32,power_law_1.01,0.029000959396362304
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,4,32,power_law_1.01,0.02952064037322998
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,4,32,power_law_1.01,0.03067903995513916
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,4,32,power_law_1.01,0.029797120094299318
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,4,32,power_law_1.01,0.03823744058609009
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,4,32,power_law_1.01,0.04922624111175537
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,4,32,power_law_1.01,0.05299456119537353
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,4,32,power_law_1.01,0.06265471935272217
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,4,32,power_law_1.01,0.08814463615417481
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,4,32,power_law_1.01,0.02420480012893677
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,4,32,power_law_1.01,0.10246399879455566
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,4,32,power_law_1.01,0.14050944328308107
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,4,32,power_law_1.01,0.2262681579589844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,4,32,power_law_1.01,0.26313343048095705
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,4,32,power_law_1.01,0.03185408115386963
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,4,32,power_law_1.01,0.36160255432128907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,4,32,power_law_1.01,0.01961472034454346
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,4,32,power_law_1.01,0.01980288028717041
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,4,32,power_law_1.01,0.02017280101776123
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,4,32,power_law_1.01,0.02871936082839966
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,4,32,power_law_1.01,0.029433600902557373
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,4,32,power_law_1.01,0.02271872043609619
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,2,64,power_law_1.2,0.6542285156250001
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,4,32,power_law_1.01,0.030385921001434325
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,4,32,power_law_1.01,0.030261759757995603
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,4,32,power_law_1.01,0.031521279811859135
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,4,32,power_law_1.01,0.031196160316467287
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,4,32,power_law_1.01,0.03248895883560181
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,4,32,power_law_1.01,0.03423232078552246
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,4,32,power_law_1.01,0.035037438869476314
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,4,32,power_law_1.01,0.037320959568023684
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,4,32,power_law_1.01,0.03385600090026856
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,4,32,power_law_1.01,0.034200320243835444
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,4,32,power_law_1.01,0.03694463968276977
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,4,32,power_law_1.01,0.03928447961807251
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,4,32,power_law_1.01,0.04540287971496582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,4,32,power_law_1.01,0.049061121940612795
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,4,32,power_law_1.01,0.059321599006652834
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,4,32,power_law_1.01,0.07132927894592285
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,4,32,power_law_1.01,0.09595264434814453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,4,32,power_law_1.01,0.12784383773803712
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,4,32,power_law_1.01,0.17094144821166993
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,4,32,power_law_1.01,0.5276236724853516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,4,32,power_law_1.01,0.2665651130676269
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,4,32,power_law_1.2,0.016710400581359863
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,4,32,power_law_1.2,0.016633599996566772
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,4,32,power_law_1.2,0.016936960220336913
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,4,32,power_law_1.01,0.35488769531250003
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,4,32,power_law_1.2,0.022451200485229493
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,4,32,power_law_1.01,0.4596851348876953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,4,32,power_law_1.2,0.02348927974700928
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,4,32,power_law_1.2,0.024410879611968993
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,4,32,power_law_1.2,0.025913600921630857
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,4,32,power_law_1.2,0.026316800117492677
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,4,32,power_law_1.2,0.027135999202728273
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,4,32,power_law_1.2,0.030316801071166993
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,4,32,power_law_1.2,0.030510079860687257
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,4,32,power_law_1.2,0.029702401161193846
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,4,32,power_law_1.2,0.02972800016403198
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,4,32,power_law_1.2,0.031031041145324706
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,4,32,power_law_1.2,0.04022272109985352
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,4,32,power_law_1.2,0.055233278274536134
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,4,32,power_law_1.2,0.06050943851470947
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,4,32,power_law_1.2,0.08809599876403809
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,2,64,power_law_1.2,0.5077260971069336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,4,32,power_law_1.2,0.10255871772766115
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,4,32,power_law_1.2,0.11736191749572752
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,4,32,power_law_1.2,0.1876582336425781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,4,32,power_law_1.2,0.023151359558105468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,4,32,power_law_1.2,0.01988479971885681
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,4,32,power_law_1.2,0.029256958961486817
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,4,32,power_law_1.2,0.020101120471954344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,4,32,power_law_1.2,0.019901440143585206
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,4,32,power_law_1.2,0.02894079923629761
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,4,32,power_law_1.2,0.02934655904769897
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,4,32,power_law_1.2,0.029982719421386715
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,4,32,power_law_1.2,0.030795519351959226
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,4,32,power_law_1.2,0.03175935983657837
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,4,32,power_law_1.2,0.031554560661315914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,4,32,power_law_1.2,0.03273216009140015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,4,32,power_law_1.2,0.03450367927551269
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,4,32,power_law_1.2,0.2670886421203613
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,4,32,power_law_1.2,0.036559360027313234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,4,32,power_law_1.2,0.03512320041656494
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,4,32,power_law_1.2,0.034410240650177004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,4,32,power_law_1.2,0.036477439403533936
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,4,32,power_law_1.2,0.036677119731903074
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,4,32,power_law_1.2,0.039622399806976315
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,4,32,power_law_1.2,0.04817024230957031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,4,32,power_law_1.2,0.05384575843811035
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,4,32,power_law_1.2,0.06648191928863526
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,4,32,power_law_1.2,0.08560895919799805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,4,32,power_law_1.2,0.12828415870666504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,4,32,power_law_1.2,0.17379711151123048
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,4,32,power_law_1.2,0.34120960235595704
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,1,128,balanced,0.02626944065093994
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,1,128,balanced,0.02707200050354004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,1,128,balanced,0.02625024080276489
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,1,128,balanced,0.027020800113677978
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,1,128,balanced,0.02622720003128052
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,1,128,balanced,0.04648575782775879
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,1,128,balanced,0.04740608215332031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,1,128,balanced,0.04727935791015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,4,32,power_law_1.2,0.25010816574096684
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,1,128,balanced,0.04666880130767822
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,1,128,balanced,0.04611839771270752
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,1,128,balanced,0.04887296199798584
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,1,128,balanced,0.05329792022705078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,1,128,balanced,0.0480076789855957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,2,64,power_law_1.2,0.6996607971191406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,1,128,balanced,0.05390592098236084
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,1,128,balanced,0.07111551761627197
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,1,128,balanced,0.06782847881317139
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,1,128,balanced,0.07353216171264648
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,1,128,balanced,0.07899263858795166
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,1,128,balanced,0.09137920379638673
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,1,128,balanced,0.08771583557128906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,1,128,balanced,0.09304320335388183
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,1,128,balanced,0.09864831924438476
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,1,128,balanced,0.11449600219726563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,1,128,balanced,0.12649472236633302
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,1,128,balanced,0.18072448730468751
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,1,128,balanced,0.2252134323120117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,1,128,balanced,0.026749439239501953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,1,128,balanced,0.06673408031463623
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,1,128,balanced,0.026594560146331787
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,1,128,balanced,0.02662015914916992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,1,128,balanced,0.026677761077880856
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,1,128,balanced,0.026598401069641113
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,1,128,balanced,0.04353919982910156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,1,128,balanced,0.04360832214355469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,1,128,balanced,0.04360320091247559
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,1,128,balanced,0.04442624092102051
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,1,128,balanced,0.0441318416595459
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,1,128,balanced,0.04409599781036377
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,1,128,balanced,0.0456166410446167
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,1,128,balanced,0.04456575870513916
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,1,128,balanced,0.045228800773620605
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,1,128,balanced,0.04822527885437012
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,1,128,balanced,0.04610047817230224
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,1,128,balanced,0.04719103813171387
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,1,128,balanced,0.047918081283569336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,1,128,balanced,0.04994944095611572
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,1,128,balanced,0.05715199947357178
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,1,128,balanced,0.05299327850341797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,1,128,balanced,0.06072319984436035
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,1,128,balanced,0.06530303955078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,1,128,balanced,0.0734169578552246
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,1,128,balanced,0.08344832420349121
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,1,128,balanced,0.10393600463867188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,1,128,balanced,0.13063551902770995
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,1,128,power_law_1.01,0.026219520568847653
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,1,128,power_law_1.01,0.026208000183105467
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,1,128,power_law_1.01,0.05291647911071777
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,1,128,power_law_1.01,0.05636096000671387
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,1,128,power_law_1.01,0.04588672161102295
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,1,128,power_law_1.01,0.05994368076324463
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,1,128,power_law_1.01,0.048793601989746097
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,4,32,power_law_1.2,0.5295910263061524
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,1,128,power_law_1.01,0.05159167766571045
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,1,128,power_law_1.01,0.07455232143402099
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,1,128,power_law_1.01,0.07623551845550537
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,1,128,power_law_1.01,0.06660223960876464
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,4,32,power_law_1.2,0.33220481872558594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,1,128,power_law_1.01,0.08076416015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,1,128,power_law_1.01,0.08885503768920898
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,1,128,power_law_1.01,0.09347968101501465
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,1,128,power_law_1.01,0.09535231590270996
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,1,128,power_law_1.01,0.09318911552429199
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,1,128,power_law_1.01,0.10644351959228515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,1,128,power_law_1.01,0.11853568077087402
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,1,128,power_law_1.01,0.170696964263916
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,1,128,power_law_1.01,0.21390848159790038
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,1,128,power_law_1.01,0.2770892715454102
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,1,128,power_law_1.01,0.07789824008941651
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,1,128,power_law_1.01,0.3806719970703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,1,128,power_law_1.01,0.6047948837280274
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,1,128,power_law_1.01,0.026908159255981445
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,1,128,power_law_1.01,0.8935807800292969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,1,128,power_law_1.01,0.026851840019226074
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,1,128,power_law_1.01,0.026549758911132815
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,1,128,power_law_1.01,0.04392704010009766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,1,128,power_law_1.01,1.2998477172851561
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,1,128,power_law_1.01,0.04395135879516602
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,1,128,power_law_1.01,0.04450943946838379
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,1,128,power_law_1.01,0.04449024200439453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,1,128,power_law_1.01,0.045391359329223634
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,4,32,power_law_1.2,0.4941427230834961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,1,128,power_law_1.01,0.04773375988006592
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,1,128,power_law_1.01,1.724226531982422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,1,128,power_law_1.01,0.04700160026550293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,1,128,power_law_1.01,0.04832255840301514
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,1,128,power_law_1.01,0.04868351936340332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,1,128,power_law_1.01,0.055107841491699214
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,1,128,power_law_1.01,0.05583615779876709
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,1,128,power_law_1.01,0.05610496044158936
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,1,128,power_law_1.01,0.06165887832641601
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,1,128,power_law_1.01,0.06733952045440675
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,1,128,power_law_1.01,0.07387775897979736
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,1,128,power_law_1.01,0.09415936470031738
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,1,128,power_law_1.01,0.11987711906433104
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,1,128,power_law_1.01,0.16782079696655275
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,1,128,power_law_1.01,0.22470272064208982
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,1,128,power_law_1.01,0.3598438262939453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,1,128,power_law_1.01,0.45347839355468744
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,1,128,power_law_1.2,0.026220800876617434
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,1,128,power_law_1.2,0.02600831985473633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,1,128,power_law_1.2,0.0459827184677124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,1,128,power_law_1.01,0.7525593566894531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,1,128,power_law_1.2,0.04673535823822021
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,1,128,power_law_1.2,0.045954561233520506
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,1,128,power_law_1.2,0.06514431953430175
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,1,128,power_law_1.2,0.07141119956970214
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,1,128,power_law_1.01,0.9716953277587891
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,1,128,power_law_1.2,0.06703872203826904
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,1,128,power_law_1.2,0.07760000228881836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,1,128,power_law_1.2,0.07756415843963624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,1,128,power_law_1.2,0.08467583656311035
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,1,128,power_law_1.2,0.09445376396179199
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,1,128,power_law_1.2,0.09089535713195801
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,1,128,power_law_1.2,0.09309696197509766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,1,128,power_law_1.2,0.10410495758056641
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,1,128,power_law_1.2,0.12449151992797851
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,1,128,power_law_1.2,0.05688960075378417
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,1,128,power_law_1.2,0.17325056076049805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,1,128,power_law_1.2,0.22515968322753904
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,1,128,power_law_1.2,0.33030654907226564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,1,128,power_law_1.2,0.5055449676513672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,1,128,power_law_1.2,0.0860966396331787
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,1,128,power_law_1.01,0.047886080741882324
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,1,128,power_law_1.2,0.026791679859161376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,1,128,power_law_1.2,0.9187596893310547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,1,128,power_law_1.2,0.026863360404968263
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,1,128,power_law_1.2,0.04328832149505615
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,1,128,power_law_1.2,0.043919358253479
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,1,128,power_law_1.2,0.0442137622833252
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,1,128,power_law_1.2,0.6673023986816407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,1,128,power_law_1.2,0.04499199867248535
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,1,128,power_law_1.2,0.04533887863159179
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,1,128,power_law_1.2,0.0463756799697876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,1,128,power_law_1.2,0.047064318656921386
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,1,128,power_law_1.2,0.0482047986984253
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,1,128,power_law_1.2,0.051847681999206544
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,1,128,power_law_1.2,0.05601535797119141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,1,128,power_law_1.2,0.06032256126403809
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,1,128,power_law_1.2,0.057820158004760744
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,1,128,power_law_1.2,0.04459392070770264
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,1,128,power_law_1.2,0.06202879905700683
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,1,128,power_law_1.2,0.06495103836059571
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,1,128,power_law_1.2,0.07709311962127685
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,1,128,power_law_1.2,0.09383296012878418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,1,128,power_law_1.2,0.14689279556274415
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,1,128,power_law_1.2,0.17129087448120117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,1,128,power_law_1.2,0.2631398391723633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,1,128,power_law_1.2,2.1032794189453123
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,1,128,power_law_1.2,0.382737922668457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,1,128,power_law_1.2,0.579546890258789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,2,64,balanced,0.0234879994392395
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,1,128,power_law_1.2,1.3789144897460939
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,2,64,balanced,0.02301055908203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,2,64,balanced,0.022909440994262696
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,1,128,power_law_1.2,1.1117298889160157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,2,64,balanced,0.026062080860137938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,2,64,balanced,0.04663680076599121
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,2,64,balanced,0.049557762145996095
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,2,64,balanced,0.023470079898834227
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,2,64,balanced,0.05506303787231446
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,2,64,balanced,0.0570854377746582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,2,64,balanced,0.0409446382522583
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,2,64,balanced,0.0532748794555664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,2,64,balanced,0.05922560214996338
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,2,64,balanced,0.0662502384185791
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,2,64,balanced,0.06306047916412354
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,2,64,balanced,0.060963840484619135
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,4,32,power_law_1.2,0.6174284744262695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,2,64,balanced,0.06239744186401367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,2,64,balanced,0.07120255947113037
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,2,64,balanced,0.08308735847473145
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,2,64,balanced,0.07766784191131591
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,2,64,balanced,0.08424703598022461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,2,64,balanced,0.08807040214538574
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,2,64,balanced,0.09456640243530273
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,2,64,balanced,0.1043289566040039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,2,64,balanced,0.1309785556793213
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,1,128,power_law_1.2,0.7739878082275391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,2,64,balanced,0.1471782398223877
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,2,64,balanced,0.21694719314575198
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,2,64,balanced,0.024975359439849854
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,2,64,balanced,0.2761727905273438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,1,128,power_law_1.2,2.794145202636719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,2,64,balanced,0.02492288112640381
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,2,64,balanced,0.024812800884246828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,2,64,balanced,0.02516864061355591
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,2,64,balanced,0.039776000976562496
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,2,64,balanced,0.040839681625366206
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,2,64,balanced,0.04105855941772461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,2,64,balanced,0.04155903816223144
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,2,64,balanced,0.04184703826904297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,2,64,balanced,0.042467842102050776
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,2,64,balanced,0.043274240493774416
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,2,64,balanced,0.04301951885223389
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,2,64,balanced,0.043831038475036624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,2,64,balanced,0.045355520248413085
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,2,64,balanced,0.050325760841369624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,2,64,balanced,0.04880512237548828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,2,64,balanced,0.05338624000549317
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,2,64,balanced,0.05863423824310303
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,2,64,balanced,0.025016319751739503
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,2,64,balanced,0.06498688220977783
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,2,64,balanced,0.07227136135101318
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,2,64,balanced,0.09125247955322266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,2,64,balanced,0.11064703941345215
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,2,64,balanced,0.0410752010345459
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,2,64,balanced,0.17979007720947265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,2,64,power_law_1.01,0.02734208106994629
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,2,64,power_law_1.01,0.031436800956726074
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,2,64,power_law_1.01,0.03227776050567627
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,2,64,power_law_1.01,0.048636159896850585
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,2,64,power_law_1.01,0.05081855773925781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,2,64,power_law_1.01,0.049373440742492676
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,2,64,power_law_1.01,0.06375679969787598
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,2,64,power_law_1.01,0.06540671825408936
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,2,64,power_law_1.01,0.0701196813583374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,4,32,power_law_1.2,0.7071040344238282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,2,64,power_law_1.01,0.06935039997100831
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,2,64,power_law_1.01,0.0736460781097412
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,2,64,power_law_1.01,0.08095359802246094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,2,64,power_law_1.01,0.08093184471130371
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,2,64,power_law_1.01,0.08198528289794922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,2,64,power_law_1.01,0.08507648468017578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,2,64,power_law_1.01,0.08751999855041503
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,2,64,balanced,0.040919041633605956
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,2,64,power_law_1.01,0.10749695777893067
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,2,64,power_law_1.01,0.1395404815673828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,2,64,power_law_1.01,0.16815616607666015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,2,64,power_law_1.01,0.21848960876464846
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,2,64,power_law_1.01,0.3201433563232422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,2,64,power_law_1.01,0.06107135772705078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,2,64,power_law_1.01,0.4787276840209961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,2,64,power_law_1.01,0.024942080974578857
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,2,64,power_law_1.01,0.02506623983383179
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,2,64,balanced,0.14460415840148927
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,2,64,power_law_1.01,0.025241599082946774
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,2,64,power_law_1.01,0.040133118629455566
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,2,64,power_law_1.01,0.9435187530517577
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,2,64,power_law_1.01,0.040501761436462405
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,2,64,power_law_1.01,0.04135551929473877
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,2,64,power_law_1.01,0.042064638137817384
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,2,64,power_law_1.01,0.04334591865539551
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,2,64,power_law_1.01,0.04309631824493408
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,2,64,power_law_1.01,0.0445363187789917
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,2,64,power_law_1.01,1.204494094848633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,2,64,power_law_1.01,0.04565760135650635
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,2,64,power_law_1.01,0.05056767940521241
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,2,64,power_law_1.01,0.05258880138397217
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,2,64,power_law_1.01,0.05319935798645019
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,2,64,power_law_1.01,0.6125324630737305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,2,64,power_law_1.01,0.05371391773223877
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,2,64,power_law_1.01,0.06305408000946044
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,2,64,power_law_1.01,0.05900159835815429
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,2,64,power_law_1.01,0.0954086399078369
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,2,64,power_law_1.01,0.09795071601867675
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,2,64,power_law_1.01,0.11719807624816894
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,2,64,power_law_1.01,0.13720191955566405
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,2,64,power_law_1.01,0.20191488265991211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,2,64,power_law_1.01,0.289172477722168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,2,64,power_law_1.01,0.043530240058898925
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,2,64,power_law_1.01,0.40487934112548823
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,2,64,power_law_1.2,0.02389375925064087
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,2,64,power_law_1.2,0.029139199256896973
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,2,64,power_law_1.2,0.03551232099533081
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,2,64,power_law_1.2,0.04057472229003906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,2,64,power_law_1.2,0.04844287872314453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,2,64,power_law_1.01,0.7495935821533204
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,2,64,power_law_1.01,0.07246335983276367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,2,64,power_law_1.2,0.04968192100524903
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,2,64,power_law_1.2,0.06344192028045655
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,2,64,power_law_1.2,0.06922751903533936
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,2,64,power_law_1.2,0.07057151794433594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,2,64,power_law_1.2,0.07392384052276611
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,2,64,power_law_1.2,0.07775231838226318
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,2,64,power_law_1.2,0.0816051197052002
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,2,64,power_law_1.2,0.08324864387512207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,2,64,power_law_1.2,0.08205951690673828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,2,64,power_law_1.2,0.08506112098693848
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,2,64,power_law_1.2,0.0926796817779541
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,2,64,power_law_1.2,0.11291007995605469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,2,64,power_law_1.2,0.12928383827209472
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,2,64,power_law_1.2,0.17261695861816406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,2,64,power_law_1.01,0.6447206115722656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,2,64,power_law_1.2,0.24917247772216794
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,2,64,power_law_1.2,0.33133056640624997
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,2,64,power_law_1.2,0.06953983783721925
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,2,64,power_law_1.2,0.024664320945739747
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,2,64,power_law_1.2,0.02550015926361084
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,2,64,power_law_1.2,0.03995520114898682
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,2,64,power_law_1.2,0.04081151962280273
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,2,64,power_law_1.2,0.0409113597869873
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,2,64,power_law_1.2,0.0425113582611084
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,2,64,power_law_1.2,0.04206592082977295
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,2,64,power_law_1.2,1.3427635192871095
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,2,64,power_law_1.2,0.04412415981292724
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,2,64,power_law_1.2,0.043162879943847654
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,2,64,power_law_1.2,0.04413440227508545
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,2,64,power_law_1.2,0.049283838272094725
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,2,64,power_law_1.2,0.05383679866790771
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,2,64,power_law_1.2,0.054963197708129886
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,2,64,power_law_1.2,0.055010561943054204
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,2,64,power_law_1.2,0.05747968196868897
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,2,64,power_law_1.2,0.05971199989318847
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,2,64,power_law_1.2,0.06876543998718262
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,2,64,power_law_1.2,0.451146240234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,2,64,power_law_1.2,0.08678272247314453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,2,64,power_law_1.2,0.1276863956451416
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,2,64,power_law_1.2,0.04743680000305176
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,2,64,power_law_1.2,0.2076326370239258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,2,64,power_law_1.2,0.6750361633300781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,2,64,power_law_1.2,0.30780031204223635
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,2,64,power_law_1.2,0.15156991958618166
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,2,64,power_law_1.2,0.8770803070068359
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,4,32,balanced,0.0216921591758728
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,4,32,balanced,0.02201215982437134
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,4,32,balanced,0.0217305588722229
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,4,32,balanced,0.022908160686492922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,4,32,balanced,0.024515841007232666
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,4,32,balanced,0.048636159896850585
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,4,32,balanced,0.053523201942443845
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,4,32,balanced,0.0504640007019043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,1,128,power_law_1.2,1.597276153564453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,4,32,balanced,0.05697023868560791
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,4,32,balanced,0.05974143981933594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,4,32,balanced,0.059953918457031255
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,4,32,balanced,0.05280640125274658
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,2,64,power_law_1.2,0.9245875549316406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,4,32,balanced,0.062497282028198244
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,4,32,balanced,0.06924543857574464
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,4,32,balanced,0.07470848083496093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,4,32,balanced,0.07293568134307862
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,4,32,balanced,0.07818111896514893
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,4,32,balanced,0.07903232097625731
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,4,32,balanced,0.08286720275878906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,4,32,balanced,0.08912256240844726
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,4,32,balanced,0.09575296401977539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,4,32,balanced,0.11389439582824708
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,4,32,balanced,0.1287104034423828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,4,32,balanced,0.16779520034790038
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,4,32,balanced,0.1987071990966797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,4,32,balanced,0.02375296115875244
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,4,32,balanced,0.023635199069976805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,4,32,balanced,0.289621753692627
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,4,32,balanced,0.3742655944824219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,4,32,balanced,0.02371840000152588
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,4,32,balanced,0.02401024103164673
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,4,32,balanced,0.02566015958786011
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,4,32,balanced,0.03857919931411743
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,4,32,balanced,0.03980671882629395
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,4,32,balanced,0.04032896041870117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,4,32,balanced,0.04054399967193604
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,4,32,balanced,0.04138751983642578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,4,32,balanced,0.04176127910614014
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,4,32,balanced,0.0440012788772583
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,4,32,balanced,0.04610047817230224
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,4,32,balanced,0.04806399822235107
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,2,64,power_law_1.2,0.48360446929931644
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,4,32,balanced,0.04536448001861572
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,4,32,balanced,0.05128704071044922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,4,32,balanced,0.05485439777374268
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,4,32,balanced,0.03928191900253296
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,4,32,balanced,0.06031744003295898
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,4,32,balanced,0.0674124813079834
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,4,32,balanced,0.08764287948608399
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,4,32,balanced,0.0416755199432373
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,4,32,balanced,0.10098815917968751
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,4,32,balanced,0.13249664306640624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,4,32,balanced,0.1640243148803711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,4,32,balanced,0.2235968017578125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,4,32,power_law_1.01,0.02476032018661499
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,4,32,balanced,0.2830105590820312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,4,32,power_law_1.01,0.028757760524749754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,4,32,power_law_1.01,0.031449599266052244
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,4,32,power_law_1.01,0.044211201667785645
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,4,32,power_law_1.01,0.04954239845275879
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,4,32,power_law_1.01,0.05415808200836182
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,2,64,power_law_1.2,0.5966719818115235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,4,32,power_law_1.01,0.06134016036987304
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,4,32,power_law_1.01,0.06676608085632324
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,4,32,power_law_1.01,0.06486783981323242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,4,32,power_law_1.01,0.07081600189208984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,4,32,power_law_1.01,0.07427584171295167
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,4,32,power_law_1.01,0.08024064064025879
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,4,32,power_law_1.01,0.08134016036987304
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,4,32,power_law_1.01,0.07840127944946289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,4,32,power_law_1.01,0.08123647689819335
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,4,32,power_law_1.01,0.08582271575927734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,4,32,power_law_1.01,0.09763839721679687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,4,32,power_law_1.01,0.11026816368103028
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,4,32,power_law_1.01,0.1750271987915039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,4,32,power_law_1.01,0.22745855331420897
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,4,32,power_law_1.01,0.27747072219848634
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,4,32,power_law_1.01,0.06115839958190918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,4,32,power_law_1.01,0.02529151916503906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,4,32,power_law_1.01,0.554095344543457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,4,32,power_law_1.01,0.02518784046173096
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,4,32,power_law_1.01,0.12468735694885254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,4,32,power_law_1.01,0.7668006134033203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,4,32,power_law_1.01,0.038257920742034913
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,4,32,power_law_1.01,0.039188480377197264
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,4,32,power_law_1.01,0.04077568054199219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,4,32,power_law_1.01,0.0397708797454834
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,4,32,power_law_1.01,0.04133120059967041
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,4,32,power_law_1.01,0.041392641067504884
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,4,32,power_law_1.01,0.0427455997467041
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,4,32,power_law_1.01,0.41971584320068356
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,4,32,power_law_1.01,0.04674943923950196
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,4,32,power_law_1.01,0.04909183979034424
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,4,32,power_law_1.01,0.05435904026031494
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,4,32,power_law_1.01,0.052439041137695305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,4,32,power_law_1.01,0.055823359489440914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,4,32,power_law_1.01,0.02559231996536255
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,4,32,power_law_1.01,0.06824575901031495
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,2,64,power_law_1.2,2.0112728881835937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,4,32,power_law_1.01,0.09463040351867676
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,4,32,power_law_1.01,1.0382259368896485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,4,32,power_law_1.01,0.12493184089660644
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,4,32,power_law_1.01,0.16936832427978515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,4,32,power_law_1.01,0.20070144653320315
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,4,32,power_law_1.01,0.04470016002655029
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,4,32,power_law_1.01,0.29823488235473633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,4,32,power_law_1.2,0.03215872049331665
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,4,32,power_law_1.01,0.43202816009521483
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,4,32,power_law_1.2,0.026545920372009278
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,4,32,power_law_1.2,0.031037440299987794
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,4,32,power_law_1.01,0.553177604675293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,4,32,power_law_1.01,0.08209664344787597
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,4,32,power_law_1.2,0.043677439689636235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,4,32,power_law_1.2,0.050367999076843264
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,4,32,power_law_1.2,0.059020800590515135
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,4,32,power_law_1.2,0.05958655834197998
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,4,32,power_law_1.2,0.06174079895019531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,4,32,power_law_1.01,0.8075161743164063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,4,32,power_law_1.2,0.06889599800109864
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,4,32,power_law_1.2,0.07371520042419434
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,4,32,power_law_1.2,0.07921664237976074
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,4,32,power_law_1.2,0.081212158203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,4,32,power_law_1.2,0.0786572790145874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,4,32,power_law_1.2,0.08102272033691407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,4,32,power_law_1.2,0.08243712425231933
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,4,32,power_law_1.2,0.08498432159423827
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,4,32,power_law_1.2,0.0966374397277832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,4,32,power_law_1.2,0.12089216232299804
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,4,32,power_law_1.2,0.14426112174987793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,4,32,power_law_1.2,0.2067545509338379
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,4,32,power_law_1.2,0.2556403160095215
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,4,32,power_law_1.2,0.048060159683227535
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,4,32,power_law_1.2,0.02524415969848633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,4,32,power_law_1.2,0.025107200145721438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,4,32,power_law_1.2,0.02534271955490112
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,4,32,power_law_1.01,0.05946623802185058
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,4,32,power_law_1.2,0.04047872066497803
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,4,32,power_law_1.2,0.33626495361328124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,4,32,power_law_1.2,0.041541118621826176
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,4,32,power_law_1.2,0.040816640853881835
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,4,32,power_law_1.2,0.04127488136291504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,4,32,power_law_1.2,0.03821568012237549
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,4,32,power_law_1.2,0.04203904151916504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,4,32,power_law_1.2,0.04227712154388428
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,4,32,power_law_1.2,0.050140161514282224
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,4,32,power_law_1.2,0.05014272212982178
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,4,32,power_law_1.2,0.05174528121948242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,4,32,power_law_1.2,0.5583014297485351
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,4,32,power_law_1.2,0.05172863960266113
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,4,32,power_law_1.2,0.053867521286010744
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,4,32,power_law_1.2,0.05557375907897949
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,4,32,power_law_1.2,0.05661439895629883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,4,32,power_law_1.2,0.06821119785308838
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,4,32,power_law_1.2,0.08372223854064942
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,4,32,power_law_1.2,0.11639039993286134
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,4,32,power_law_1.2,0.14146047592163086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,4,32,power_law_1.2,0.1981376075744629
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,4,32,power_law_1.2,0.26066944122314456
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,2,64,power_law_1.2,1.2306803131103516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,4,32,power_law_1.2,0.7232486724853515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,8,16,balanced,0.021388800144195558
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,8,16,balanced,0.021245439052581788
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,8,16,balanced,0.023778560161590575
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,8,16,balanced,0.03757184028625488
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,8,16,balanced,0.061813759803771975
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,4,32,power_law_1.2,1.3002584838867188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,8,16,balanced,0.06506495952606202
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,4,32,power_law_1.2,0.7995430755615234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,4,32,power_law_1.2,1.0656281280517579
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,8,16,balanced,0.06446720123291015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,8,16,balanced,0.06179967880249023
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,8,16,balanced,0.05987328052520752
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,8,16,balanced,0.07343616008758544
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,8,16,balanced,0.07645440101623535
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,8,16,balanced,0.07007616043090821
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,8,16,balanced,0.02211456060409546
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,8,16,balanced,0.07439616203308105
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,8,16,balanced,0.07694464206695557
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,8,16,balanced,0.08097023963928222
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,8,16,balanced,0.08554623603820802
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,8,16,balanced,0.09336447715759277
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,8,16,balanced,0.10846847534179686
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,8,16,balanced,0.1222771167755127
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,8,16,balanced,0.15366016387939455
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,8,16,balanced,0.1817241668701172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,8,16,balanced,0.2439142417907715
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,8,16,balanced,0.2997452735900879
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,8,16,balanced,0.023624958992004393
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,8,16,balanced,0.43968769073486325
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,8,16,balanced,0.02366976022720337
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,4,32,power_law_1.2,0.41758464813232427
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,8,16,balanced,0.06926208019256593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,8,16,balanced,0.024252159595489504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,8,16,balanced,0.5784614562988282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,8,16,balanced,0.024422399997711182
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,8,16,balanced,0.0414899206161499
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,8,16,balanced,0.0471884822845459
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,8,16,balanced,0.044846081733703615
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,8,16,balanced,0.04562431812286377
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,8,16,balanced,0.04768767833709717
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,8,16,balanced,0.05007872104644775
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,8,16,balanced,0.04850048065185547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,8,16,balanced,0.051357440948486335
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,8,16,balanced,0.052541441917419426
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,8,16,balanced,0.05623040199279785
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,8,16,balanced,0.060737280845642096
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,8,16,balanced,0.06694143772125244
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,8,16,balanced,0.08394368171691895
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,8,16,balanced,0.09712512016296386
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,8,16,balanced,0.12922752380371094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,8,16,balanced,0.15661567687988281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,8,16,balanced,0.21474559783935546
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,8,16,balanced,0.04265984058380127
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,8,16,balanced,0.27034496307373046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,8,16,balanced,0.38736385345458985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,8,16,balanced,0.051244797706604006
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,8,16,power_law_1.01,0.024264960289001464
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,8,16,power_law_1.01,0.032040960788726806
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,8,16,power_law_1.01,0.028720641136169435
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,8,16,balanced,0.4998387145996094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,8,16,power_law_1.01,0.05036159992218018
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,8,16,power_law_1.01,0.05912447929382324
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,8,16,power_law_1.01,0.06329728126525878
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,8,16,power_law_1.01,0.06566912174224854
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,8,16,power_law_1.01,0.07017216205596924
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,8,16,power_law_1.01,0.07412479877471924
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,8,16,power_law_1.01,0.07805952072143554
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,8,16,balanced,0.026245119571685793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,8,16,power_law_1.01,0.07405951976776123
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,8,16,power_law_1.01,0.07467904090881347
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,8,16,power_law_1.01,0.07821184158325195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,8,16,power_law_1.01,0.08158080101013183
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,8,16,power_law_1.01,0.08636672019958495
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,8,16,power_law_1.01,0.09716480255126954
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,8,16,power_law_1.01,0.10791168212890626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,8,16,power_law_1.01,0.12111104011535645
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,8,16,power_law_1.01,0.031868159770965576
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,8,16,power_law_1.01,0.06420224189758302
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,8,16,power_law_1.01,0.22782848358154295
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,8,16,power_law_1.01,0.25043840408325196
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,8,16,power_law_1.01,0.4169881439208984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,4,32,power_law_1.2,0.5506700897216797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,8,16,power_law_1.01,0.46852352142333986
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,8,16,power_law_1.01,0.025343999862670896
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,8,16,power_law_1.01,0.025890560150146485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,8,16,power_law_1.01,0.7539673614501953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,8,16,power_law_1.01,0.02616192102432251
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,8,16,power_law_1.01,0.041023998260498046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,8,16,power_law_1.01,0.16162431716918946
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,8,16,power_law_1.01,0.04412928104400635
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,8,16,power_law_1.01,0.04360832214355469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,8,16,power_law_1.01,0.045274882316589354
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,8,16,power_law_1.01,0.045457921028137206
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,8,16,power_law_1.01,0.047576317787170405
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,8,16,power_law_1.01,0.05026048183441162
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,8,16,power_law_1.01,0.05158527851104736
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,8,16,power_law_1.01,0.054640641212463384
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,8,16,power_law_1.01,0.05256063938140869
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,8,16,power_law_1.01,0.05536384105682372
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,8,16,power_law_1.01,0.057752318382263176
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,8,16,power_law_1.01,0.06383999824523925
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,8,16,power_law_1.01,0.025333759784698484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,8,16,power_law_1.01,0.07118207931518554
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,8,16,power_law_1.01,0.08604288101196289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,8,16,power_law_1.01,0.11563776016235353
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,8,16,power_law_1.01,0.12852224349975586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,8,16,power_law_1.01,0.18429183959960938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,8,16,power_law_1.01,0.23750783920288088
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,8,16,power_law_1.01,0.3221043014526367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,8,16,power_law_1.01,1.0653337860107421
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,8,16,power_law_1.2,0.026607360839843754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,8,16,power_law_1.01,0.48757377624511716
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,8,16,power_law_1.2,0.03297791957855224
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,8,16,power_law_1.2,0.03469568014144898
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,8,16,power_law_1.01,0.6306752014160156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,8,16,power_law_1.2,0.05601024150848388
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,8,16,power_law_1.2,0.05008512020111085
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,8,16,power_law_1.2,0.06482431888580323
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,8,16,power_law_1.2,0.061610240936279294
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,8,16,power_law_1.2,0.06707839965820313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,8,16,power_law_1.01,0.8784678649902343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,8,16,power_law_1.2,0.07445248126983642
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,8,16,power_law_1.2,0.07891071796417236
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,8,16,power_law_1.2,0.07584640026092529
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,8,16,power_law_1.2,0.082608642578125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,8,16,power_law_1.2,0.08096639633178711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,8,16,power_law_1.2,0.08613632202148438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,8,16,power_law_1.2,0.09988991737365722
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,8,16,power_law_1.2,0.11610752105712892
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,8,16,power_law_1.2,0.14801024436950683
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,8,16,power_law_1.2,0.03249151945114136
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,8,16,power_law_1.2,0.1797145652770996
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,8,16,power_law_1.2,0.2423539161682129
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,8,16,power_law_1.2,0.06878079891204833
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,8,16,power_law_1.2,0.07578112125396728
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,8,16,power_law_1.2,0.02539007902145386
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,8,16,power_law_1.2,0.025483520030975343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,8,16,power_law_1.2,0.4655295944213867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,8,16,power_law_1.2,0.025822720527648925
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,8,16,power_law_1.2,0.026575360298156735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,8,16,power_law_1.2,0.811072006225586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,8,16,power_law_1.2,0.03908735990524292
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,8,16,power_law_1.2,0.04178304195404053
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,8,16,power_law_1.2,0.045112318992614746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,8,16,power_law_1.2,0.044247040748596186
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,8,16,power_law_1.2,0.046073598861694334
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,8,16,power_law_1.2,0.05272960186004638
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,8,16,power_law_1.2,0.05834623813629151
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,8,16,power_law_1.2,0.27637887954711915
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,8,16,power_law_1.2,0.05520256042480469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,8,16,power_law_1.2,0.05311744213104248
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,8,16,power_law_1.2,0.05711743831634521
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,8,16,power_law_1.2,0.059261441230773926
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,8,16,power_law_1.2,0.06051583766937256
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,8,16,power_law_1.2,0.07720831871032716
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,8,16,power_law_1.2,0.08576000213623047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,8,16,power_law_1.2,0.12067839622497559
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,8,16,power_law_1.2,0.13869183540344238
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,8,16,power_law_1.2,0.20300672531127928
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,8,16,power_law_1.2,0.047513599395751956
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,8,16,power_law_1.2,0.38588417053222657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,8,16,power_law_1.2,0.6023372650146485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,8,16,power_law_1.2,0.2550899124145508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,16,8,balanced,0.02153599977493286
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,16,8,balanced,0.022421760559082033
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,16,8,balanced,0.027045118808746337
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,16,8,balanced,0.041277441978454586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,16,8,balanced,0.06595583915710448
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,16,8,balanced,0.06778624057769775
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,16,8,balanced,0.07065728187561035
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,16,8,balanced,0.07310719966888428
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,16,8,balanced,0.07811583995819091
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,16,8,balanced,0.07086336135864257
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,16,8,balanced,0.07395455837249756
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,16,8,balanced,0.07611135959625244
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,16,8,balanced,0.08035455703735352
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,16,8,balanced,0.08219648361206054
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,16,8,balanced,0.023840000629425047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,16,8,balanced,0.08502528190612793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,16,8,balanced,0.09185664176940918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,16,8,balanced,0.10632831573486327
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,16,8,balanced,0.11933055877685547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,16,8,balanced,0.14899840354919433
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,16,8,balanced,0.1760972785949707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,16,8,balanced,0.23240320205688478
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,16,8,balanced,0.2859929656982422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,8,16,power_law_1.2,0.5187660980224609
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,16,8,balanced,0.3967820739746094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,4,32,power_law_1.2,1.1098099517822266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,16,8,balanced,0.02376703977584839
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,16,8,balanced,0.506355209350586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,16,8,balanced,0.024223999977111815
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,16,8,balanced,0.024450559616088868
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,16,8,balanced,0.026604800224304198
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,16,8,balanced,0.03078144073486328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,16,8,balanced,0.04819200038909912
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,16,8,balanced,0.048427519798278806
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,16,8,balanced,0.9988620758056641
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,16,8,balanced,0.05124095916748047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,16,8,balanced,0.05284607887268067
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,16,8,balanced,0.05170944213867188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,16,8,balanced,0.05524352073669434
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,16,8,balanced,0.05450751781463623
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,16,8,balanced,0.0558681583404541
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,16,8,balanced,0.05939199924468994
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,16,8,balanced,0.060888319015502936
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,16,8,balanced,0.06297215938568115
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,16,8,balanced,0.06978303909301757
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,16,8,balanced,0.08542592048645019
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,16,8,balanced,0.09943296432495116
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,16,8,balanced,0.12622079849243165
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,16,8,balanced,0.15229184150695801
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,16,8,balanced,0.21312255859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,16,8,balanced,0.2612940788269043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,16,8,balanced,0.37472896575927733
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,16,8,balanced,0.4827596664428711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,16,8,power_law_1.01,0.024261119365692137
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,16,8,power_law_1.01,0.028199679851531982
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,16,8,balanced,0.7042150115966797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,16,8,power_law_1.01,0.03737855911254882
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,16,8,balanced,0.7516223907470703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,16,8,power_law_1.01,0.0385971188545227
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,16,8,power_law_1.01,0.03926271915435791
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,16,8,power_law_1.01,0.06784128189086915
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,16,8,power_law_1.01,0.06623487949371339
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,16,8,power_law_1.01,0.07048831939697266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,16,8,power_law_1.01,0.07512063980102539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,16,8,power_law_1.01,0.07760640144348144
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,16,8,power_law_1.01,0.07666304111480712
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,16,8,power_law_1.01,0.0741107177734375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,16,8,power_law_1.01,0.07823999881744384
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,16,8,power_law_1.01,0.07967743873596192
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,16,8,power_law_1.01,0.08205951690673828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,16,8,power_law_1.01,0.0887769603729248
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,16,8,power_law_1.01,0.10300031661987305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,16,8,power_law_1.01,0.1152524757385254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,16,8,power_law_1.01,0.14339584350585938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,16,8,power_law_1.01,0.06460671901702882
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,16,8,power_law_1.01,0.17358591079711913
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,16,8,power_law_1.01,0.21057920455932616
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,16,8,power_law_1.01,0.2587366485595703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,16,8,power_law_1.01,0.37682174682617187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,16,8,power_law_1.01,0.4514009475708008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,16,8,power_law_1.01,0.02517503976821899
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,16,8,power_law_1.01,0.026699519157409667
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,16,8,power_law_1.01,0.7435993957519531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,16,8,power_law_1.01,0.028053760528564453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,16,8,balanced,0.9264832305908204
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,16,8,power_law_1.01,0.029515519142150882
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,16,8,power_law_1.01,0.04677120208740234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,16,8,power_law_1.01,0.9573619079589843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,16,8,power_law_1.01,0.04964735984802246
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,16,8,power_law_1.01,0.051637759208679204
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,16,8,power_law_1.01,0.052387838363647465
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,16,8,power_law_1.01,0.05994495868682861
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,16,8,power_law_1.01,0.054321918487548825
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,16,8,power_law_1.01,0.055877118110656744
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,16,8,power_law_1.01,0.057925119400024414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,16,8,power_law_1.01,0.05907584190368652
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,16,8,power_law_1.01,0.06148223876953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,16,8,power_law_1.01,0.06874112129211427
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,16,8,power_law_1.01,0.08230912208557128
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,16,8,power_law_1.01,0.10040703773498535
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,16,8,power_law_1.01,0.029836800098419193
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,16,8,power_law_1.01,0.11989119529724121
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,16,8,power_law_1.01,0.1602521514892578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,16,8,power_law_1.01,0.21264255523681638
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,16,8,power_law_1.01,0.26116735458374024
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,16,8,power_law_1.01,0.36758144378662105
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,8,16,power_law_1.2,0.8051750183105468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,16,8,power_law_1.2,0.023953919410705564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,16,8,power_law_1.2,0.03075968027114868
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,16,8,power_law_1.2,0.03683583974838257
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,16,8,power_law_1.01,0.7521830749511719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,16,8,power_law_1.01,0.047528958320617674
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,16,8,power_law_1.2,0.03859967947006225
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,16,8,power_law_1.2,0.040181760787963865
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,16,8,power_law_1.2,0.06832640171051026
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,16,8,power_law_1.2,0.06459263801574708
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,16,8,power_law_1.2,0.06851456165313721
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,16,8,power_law_1.2,0.07135488033294678
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,16,8,power_law_1.2,0.0737446403503418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,16,8,power_law_1.2,0.07839360237121581
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,16,8,power_law_1.2,0.07310719966888428
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,16,8,power_law_1.2,0.07590271949768065
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,16,8,power_law_1.01,0.4791616058349609
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,16,8,power_law_1.2,0.07744383811950684
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,16,8,power_law_1.2,0.08133503913879395
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,16,8,power_law_1.2,0.08469120025634766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,16,8,power_law_1.2,0.09342464447021484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,16,8,power_law_1.2,0.10485119819641113
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,16,8,power_law_1.2,0.11768832206726074
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,16,8,power_law_1.2,0.14640640258789062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,16,8,power_law_1.2,0.1749772834777832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,16,8,power_law_1.2,0.23025279998779297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,16,8,power_law_1.2,0.2937689590454101
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,16,8,power_law_1.01,0.9248921966552734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,16,8,power_law_1.2,0.025176320075988766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,16,8,power_law_1.2,0.025784320831298828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,16,8,power_law_1.2,0.027831039428710937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,16,8,power_law_1.2,0.027631359100341795
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,16,8,power_law_1.2,0.029521920680999757
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,16,8,power_law_1.2,0.046337280273437496
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,16,8,power_law_1.2,0.047930879592895506
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,16,8,power_law_1.2,0.049548802375793455
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,16,8,power_law_1.2,0.05142144203186035
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,16,8,power_law_1.2,0.052876801490783686
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,16,8,power_law_1.2,0.057021441459655764
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,16,8,power_law_1.2,0.05620480060577393
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,16,8,power_law_1.2,0.05506688117980957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,16,8,power_law_1.2,0.05812736034393311
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,16,8,power_law_1.2,0.06111104011535644
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,16,8,power_law_1.2,0.06425471782684326
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,16,8,power_law_1.2,0.4254451370239257
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,16,8,power_law_1.2,0.07184256076812744
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,16,8,power_law_1.2,0.0878105640411377
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,16,8,power_law_1.2,0.09668736457824707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,16,8,power_law_1.2,0.13394687652587892
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,16,8,power_law_1.2,0.16469247817993166
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,16,8,power_law_1.2,0.40392574310302737
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,16,8,power_law_1.2,0.22386047363281253
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,16,8,power_law_1.2,0.5673587036132812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,16,8,power_law_1.2,0.5353779220581055
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,32,4,balanced,0.0229913592338562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,32,4,balanced,0.023873279094696047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,32,4,balanced,0.026103041172027587
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,32,4,balanced,0.02956928014755249
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,32,4,balanced,0.04519040107727051
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,16,8,power_law_1.2,0.2800972747802734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,32,4,balanced,0.04800896167755127
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,32,4,balanced,0.07156479835510254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,32,4,balanced,0.07198463916778565
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,32,4,balanced,0.07631231784820557
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,32,4,balanced,0.07981696128845214
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,32,4,balanced,0.07251455783843994
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,32,4,balanced,0.07572991847991943
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,32,4,balanced,0.07807104110717773
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,32,4,balanced,0.08146431922912598
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,32,4,balanced,0.08294400215148925
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,32,4,balanced,0.08579839706420898
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,32,4,balanced,0.09161855697631835
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,32,4,balanced,0.10657664299011231
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,32,4,balanced,0.1196671962738037
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,32,4,balanced,0.15107839584350585
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,32,4,balanced,0.17682048797607422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,16,8,power_law_1.2,0.7584703826904298
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,32,4,balanced,0.23271936416625977
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,32,4,balanced,0.28824960708618164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,32,4,balanced,0.3990796661376953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,32,4,balanced,0.5232486343383789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,32,4,balanced,0.02482815980911255
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,8,16,power_law_1.2,1.0520012664794922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,32,4,balanced,0.7934937286376953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,32,4,balanced,0.027159039974212644
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,32,4,balanced,0.030928640365600585
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,32,4,balanced,1.0317938995361327
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,32,4,balanced,0.038365440368652345
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,32,4,balanced,0.040595197677612306
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,32,4,balanced,0.058490881919860835
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,32,4,balanced,0.06457727909088135
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,32,4,balanced,0.06371071815490723
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,32,4,balanced,0.0605734395980835
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,32,4,balanced,0.06316927909851075
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,32,4,balanced,0.0649228811264038
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,32,4,balanced,0.02499711990356445
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,32,4,balanced,0.06795135974884034
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,32,4,balanced,0.06899199962615968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,32,4,balanced,0.07179391860961915
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,32,4,balanced,0.09403264045715332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,32,4,balanced,0.10663552284240722
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,32,4,balanced,0.1343488025665283
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,32,4,balanced,0.06036096096038819
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,16,8,power_law_1.2,0.7258265686035157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,32,4,balanced,0.22602752685546873
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,32,4,balanced,0.2730624008178711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,32,4,balanced,0.3890995025634766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,32,4,balanced,0.07795711994171142
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,32,4,power_law_1.01,0.024225280284881592
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,32,4,balanced,0.7589746856689453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,32,4,balanced,0.1575577640533447
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,32,4,power_law_1.01,0.028163840770721437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,32,4,power_law_1.01,0.03573375940322876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,32,4,balanced,0.9936640167236328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,32,4,power_law_1.01,0.04321407794952393
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,32,4,power_law_1.01,0.050810880661010735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,32,4,balanced,0.5058943939208984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,32,4,power_law_1.01,0.06882944107055664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,32,4,power_law_1.01,0.07584896087646484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,32,4,power_law_1.01,0.07769728183746338
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,32,4,power_law_1.01,0.07881343841552735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,32,4,power_law_1.01,0.07420544147491455
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,32,4,power_law_1.01,0.07651584148406983
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,32,4,power_law_1.01,0.04077824115753174
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,32,4,power_law_1.01,0.08304639816284179
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,32,4,power_law_1.01,0.08547712326049804
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,32,4,power_law_1.01,0.09127424240112306
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,32,4,power_law_1.01,0.10889984130859376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,32,4,power_law_1.01,0.1176358413696289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,32,4,power_law_1.01,0.14964096069335936
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,32,4,power_law_1.01,0.18295936584472655
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,32,4,power_law_1.01,0.2408025550842285
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,32,4,power_law_1.01,0.08118911743164062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,32,4,power_law_1.01,0.2876249694824219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,32,4,power_law_1.01,0.07263360023498536
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,32,4,power_law_1.01,0.42302719116210935
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,32,4,power_law_1.01,0.02596479892730713
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,32,4,power_law_1.01,0.5709452819824219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,32,4,power_law_1.01,0.028488960266113285
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,32,4,power_law_1.01,0.031169281005859372
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,32,4,power_law_1.01,0.033783040046691894
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,32,4,power_law_1.01,0.03473536014556885
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,32,4,power_law_1.01,0.8754431915283203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,32,4,power_law_1.01,0.0419865608215332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,32,4,power_law_1.01,0.05737343788146972
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,32,4,power_law_1.01,0.060695037841796876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,32,4,power_law_1.01,0.0605452823638916
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,32,4,power_law_1.01,0.0652351999282837
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,32,4,power_law_1.01,0.06435200214385986
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,32,4,power_law_1.01,1.0972351837158203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,32,4,power_law_1.01,0.06193408012390137
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,32,4,power_law_1.01,0.06393983840942383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,32,4,power_law_1.01,0.06781184196472168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,32,4,power_law_1.01,0.06796160221099853
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,32,4,power_law_1.01,0.07121535778045654
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,32,4,power_law_1.01,0.07773439884185791
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,32,4,power_law_1.01,0.09601920127868653
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,32,4,power_law_1.01,0.1123481559753418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,32,4,power_law_1.01,0.14487808227539062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,32,4,power_law_1.01,0.17260927200317383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,32,4,power_law_1.01,0.2356108856201172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,32,4,power_law_1.01,0.30037376403808597
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,32,4,power_law_1.01,0.43329792022705077
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,32,4,power_law_1.2,0.024413440227508545
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,32,4,power_law_1.01,0.5621772766113281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,32,4,power_law_1.2,0.02807039976119995
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,32,4,power_law_1.2,0.03496063947677612
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,32,4,power_law_1.2,0.03995392084121704
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,32,4,power_law_1.2,0.04456575870513916
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,32,4,power_law_1.01,0.8418240356445313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,32,4,power_law_1.2,0.04777728080749512
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,8,16,power_law_1.2,1.1932083129882813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,32,4,power_law_1.2,0.06998655796051026
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,32,4,power_law_1.2,0.0721830415725708
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,32,4,power_law_1.2,0.07698304176330566
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,32,4,power_law_1.01,1.0712242889404298
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,32,4,power_law_1.2,0.08258560180664062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,32,4,power_law_1.2,0.07505407810211182
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,32,4,power_law_1.2,0.08048895835876466
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,32,4,power_law_1.2,0.08399488449096679
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,32,4,power_law_1.2,0.09367679595947266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,32,4,power_law_1.2,0.10804096221923828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,32,4,power_law_1.2,0.12132991790771483
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,32,4,power_law_1.2,0.1518937587738037
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,32,4,power_law_1.2,0.1827622413635254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,32,4,power_law_1.2,0.24423295974731446
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,32,4,power_law_1.2,0.07756415843963624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,32,4,power_law_1.2,0.07440639972686768
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,32,4,power_law_1.2,0.3005491256713867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,32,4,power_law_1.2,0.44533248901367184
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,32,4,power_law_1.2,0.08614527702331543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,32,4,power_law_1.2,0.02596735954284668
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,32,4,power_law_1.2,0.027679998874664307
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,32,4,power_law_1.2,0.029954559803009033
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,32,4,power_law_1.2,0.032948479652404786
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,32,4,power_law_1.2,0.03419136047363282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,32,4,power_law_1.2,0.039312639236450196
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,32,4,power_law_1.2,0.0578111982345581
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,32,4,power_law_1.2,0.059333119392395016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,32,4,power_law_1.2,0.06204415798187256
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,32,4,power_law_1.2,0.0661568021774292
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,32,4,power_law_1.2,0.062414078712463385
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,32,4,power_law_1.2,0.060811519622802734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,32,4,power_law_1.2,0.06392831802368164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,32,4,power_law_1.2,0.06684160232543945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,32,4,power_law_1.2,0.06923391819000244
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,32,4,power_law_1.2,0.07188735961914063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,32,4,power_law_1.2,0.0813708782196045
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,32,4,power_law_1.2,0.09672320365905762
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,32,4,power_law_1.2,0.11515392303466795
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,32,4,power_law_1.2,0.14598527908325196
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,32,4,power_law_1.2,0.1754547119140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,32,4,power_law_1.2,0.5838835144042969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,32,4,power_law_1.2,0.3066227149963379
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,32,4,power_law_1.2,0.24059007644653319
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,32,4,power_law_1.2,0.44096126556396487
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,4,32,balanced,0.02813823938369751
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,4,32,balanced,0.02857088088989258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,4,32,balanced,0.028637440204620362
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,32,4,power_law_1.2,0.5684467315673828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,4,32,balanced,0.05061503887176514
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,4,32,balanced,0.05912447929382324
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,4,32,balanced,0.10566911697387696
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,4,32,balanced,0.10461312294006349
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,4,32,balanced,0.10869759559631346
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,4,32,balanced,0.10582015991210938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,4,32,balanced,0.11178751945495606
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,4,32,balanced,0.10900480270385741
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,4,32,balanced,0.11073023796081542
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,4,32,balanced,0.11353471755981445
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,4,32,balanced,0.11652992248535157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,4,32,balanced,0.11973888397216796
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,4,32,balanced,0.1234547233581543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,4,32,balanced,0.11757823944091797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,4,32,balanced,0.12525823593139648
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,4,32,balanced,0.13160191535949708
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,4,32,balanced,0.14618495941162107
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,4,32,balanced,0.15531135559082032
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,32,4,power_law_1.2,0.8865907287597656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,4,32,balanced,0.18301311492919922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,4,32,balanced,0.21020032882690431
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,4,32,balanced,0.2808012771606445
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,4,32,balanced,0.3711641693115234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,4,32,balanced,0.5108083343505859
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,4,32,balanced,0.028714239597320557
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,4,32,balanced,0.02891263961791992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,4,32,balanced,0.6762073516845704
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,4,32,balanced,0.02978431940078735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,4,32,balanced,0.037320959568023684
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,4,32,balanced,0.056675839424133304
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,4,32,balanced,0.058172159194946296
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,4,32,balanced,0.059390721321105955
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,4,32,balanced,0.05909632205963135
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,4,32,balanced,0.06074495792388916
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,4,32,balanced,0.061537280082702636
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,4,32,balanced,0.028679680824279786
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,4,32,balanced,0.06333951950073242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,4,32,balanced,0.06703231811523438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,4,32,balanced,0.06848512172698976
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,4,32,balanced,0.07388671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,4,32,balanced,0.06713344097137451
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,4,32,balanced,0.07192063808441163
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,4,32,balanced,0.07443712234497071
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,4,32,balanced,0.0854963207244873
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,4,32,balanced,0.060798721313476564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,4,32,balanced,0.09733887672424316
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,4,32,balanced,0.12330240249633788
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,4,32,balanced,0.1431884765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,4,32,balanced,0.18974847793579103
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,4,32,balanced,0.2342412757873535
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,32,4,power_law_1.2,0.8438630676269531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,4,32,balanced,0.3194457626342774
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,4,32,power_law_1.01,0.06141439914703369
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,4,32,balanced,0.44173313140869136
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,4,32,power_law_1.01,0.061432318687438967
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,4,32,power_law_1.01,0.06159103870391845
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,4,32,power_law_1.01,0.058604798316955566
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,4,32,power_law_1.01,0.10376576423645019
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,4,32,power_law_1.01,0.10486144065856932
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,4,32,power_law_1.01,0.10994432449340821
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,4,32,power_law_1.01,0.10627200126647948
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,4,32,power_law_1.01,0.10794624328613281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,4,32,power_law_1.01,0.11402367591857909
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,4,32,power_law_1.01,0.11581952095031738
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,4,32,power_law_1.01,0.12019968032836914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,4,32,power_law_1.01,0.1243660831451416
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,4,32,power_law_1.01,0.1232755184173584
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,4,32,power_law_1.01,0.12490880012512207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,4,32,power_law_1.01,0.1370265579223633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,4,32,power_law_1.01,0.1529535961151123
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,4,32,power_law_1.01,0.1771891212463379
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,4,32,power_law_1.01,0.10246399879455566
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,4,32,power_law_1.01,0.2090777587890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,4,32,power_law_1.01,0.2851148796081543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,4,32,power_law_1.01,0.38695934295654294
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,4,32,power_law_1.01,0.49645950317382814
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,4,32,power_law_1.01,0.6597350311279297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,4,32,power_law_1.01,0.04007808208465576
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,4,32,power_law_1.01,0.8690790557861329
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,4,32,power_law_1.01,0.04078464031219482
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,4,32,power_law_1.01,0.04111231803894043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,4,32,power_law_1.01,0.04109824180603027
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,4,32,power_law_1.01,0.055505919456481936
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,4,32,power_law_1.01,1.44
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,4,32,power_law_1.01,0.05866496086120605
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,4,32,power_law_1.01,0.059470081329345705
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,4,32,power_law_1.01,0.060707840919494625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,4,32,power_law_1.01,0.06305535793304443
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,4,32,power_law_1.01,0.061966080665588376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,4,32,power_law_1.01,1.9544000244140627
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,4,32,power_law_1.01,0.06533120155334474
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,4,32,power_law_1.01,0.07203711986541747
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,4,32,power_law_1.01,0.07655424118041992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,4,32,power_law_1.01,0.07012224197387695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,4,32,power_law_1.01,0.07119359970092773
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,4,32,power_law_1.01,0.07792768001556397
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,4,32,power_law_1.01,0.09243647575378418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,4,32,power_law_1.01,0.10991616249084472
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,4,32,power_law_1.01,0.12751999855041504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,4,32,power_law_1.01,0.06921088218688964
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,4,32,power_law_1.01,0.16345087051391602
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,4,32,power_law_1.01,0.21976831436157224
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,4,32,power_law_1.01,0.3275545501708984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,4,32,power_law_1.01,0.45435009002685545
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,4,32,power_law_1.01,0.6081267166137695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,4,32,power_law_1.2,0.06167808055877686
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,16,8,power_law_1.2,1.0430643463134766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,4,32,power_law_1.01,0.8153612518310547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,4,32,power_law_1.2,0.061399040222167967
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,4,32,power_law_1.2,0.06637311935424804
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,4,32,power_law_1.2,0.10099967956542968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,4,32,power_law_1.2,0.09825152397155762
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,4,32,power_law_1.2,0.10359040260314942
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,4,32,power_law_1.2,0.10466176033020021
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,4,32,power_law_1.01,0.9617830657958985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,4,32,power_law_1.2,0.10846847534179686
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,32,4,power_law_1.2,1.1027238464355469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,4,32,power_law_1.2,0.1076531219482422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,4,32,power_law_1.2,0.10888704299926757
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,4,32,power_law_1.2,0.11594752311706542
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,4,32,power_law_1.2,0.1257868766784668
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,4,32,power_law_1.2,0.1243507194519043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,4,32,power_law_1.2,0.12107775688171388
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,4,32,power_law_1.2,0.13226367950439452
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,4,32,power_law_1.2,0.14141056060791016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,4,32,power_law_1.2,0.16819072723388673
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,4,32,power_law_1.2,0.19119232177734374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,4,32,power_law_1.2,0.33480575561523435
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,4,32,power_law_1.2,0.2654220771789551
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,4,32,power_law_1.2,0.519502067565918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,4,32,power_law_1.2,0.6283340835571289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,4,32,power_law_1.2,0.1277939224243164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,4,32,power_law_1.2,0.039888639450073246
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,4,32,power_law_1.2,0.8930944061279297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,4,32,power_law_1.2,0.03996927976608276
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,4,32,power_law_1.2,0.04169856071472168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,4,32,power_law_1.2,0.05291007995605469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,4,32,power_law_1.2,0.055659518241882325
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,4,32,power_law_1.2,0.05899775981903076
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,4,32,power_law_1.2,1.3153190612792969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,4,32,power_law_1.2,0.05932928085327148
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,4,32,power_law_1.2,0.060985598564147946
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,4,32,power_law_1.2,0.062403841018676756
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,4,32,power_law_1.2,0.06646143913269043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,4,32,power_law_1.2,0.0617907190322876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,4,32,power_law_1.2,0.06989823818206788
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,4,32,power_law_1.2,0.07419263839721679
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,4,32,power_law_1.2,0.07115903854370118
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,4,32,power_law_1.2,0.07535103797912598
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,4,32,power_law_1.2,0.08016768455505371
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,4,32,power_law_1.2,0.10038399696350098
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,4,32,power_law_1.2,0.11934592247009276
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,4,32,power_law_1.2,0.1637183952331543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,4,32,power_law_1.2,0.18889087677001953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,4,32,power_law_1.2,2.4844415283203127
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,4,32,power_law_1.2,0.26092031478881833
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,4,32,power_law_1.2,0.37049217224121095
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,4,32,power_law_1.2,0.06978303909301757
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,4,32,power_law_1.2,0.526120948791504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,8,16,balanced,0.027663359642028807
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,32,4,power_law_1.2,1.1306649780273437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,8,16,balanced,0.027743999958038328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,8,16,balanced,0.029008638858795167
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,4,32,power_law_1.2,0.7682022094726563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,8,16,balanced,0.059048957824707024
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,8,16,balanced,0.10257535934448243
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,8,16,balanced,0.10575615882873535
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,8,16,balanced,0.10873215675354005
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,8,16,balanced,0.10972543716430665
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,8,16,balanced,0.11170175552368164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,8,16,balanced,0.11519743919372558
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,8,16,balanced,0.11743231773376465
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,8,16,balanced,0.121463041305542
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,8,16,balanced,0.11494144439697265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,8,16,balanced,0.11795200347900389
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,8,16,balanced,0.12159104347229004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,8,16,balanced,0.1262476825714111
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,4,32,power_law_1.2,1.4632601928710938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,8,16,balanced,0.13890432357788085
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,8,16,balanced,0.14756223678588867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,8,16,balanced,0.1692416000366211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,8,16,balanced,0.0509657621383667
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,8,16,balanced,0.18959615707397462
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,8,16,balanced,0.27932416915893554
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,8,16,balanced,0.3786751937866211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,8,16,balanced,0.02822016000747681
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,8,16,balanced,0.5028096008300781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,8,16,balanced,0.028330240249633785
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,8,16,balanced,0.028787200450897214
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,8,16,balanced,0.7176268768310547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,16,8,power_law_1.2,0.9394303894042968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,8,16,balanced,0.03060352087020874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,8,16,balanced,0.04104063987731933
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,8,16,balanced,0.06482175827026368
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,8,16,balanced,0.9481228637695314
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,8,16,balanced,0.06560128211975098
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,8,16,balanced,0.22938880920410157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,8,16,balanced,0.06609407901763917
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,8,16,balanced,0.07005568027496338
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,8,16,balanced,0.06743807792663574
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,8,16,balanced,0.07250815868377686
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,8,16,balanced,0.07581567764282227
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,8,16,balanced,0.0693555212020874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,8,16,balanced,0.07149824142456054
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,8,16,balanced,0.07354752063751221
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,8,16,balanced,0.07636864185333252
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,8,16,balanced,0.08616448402404785
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,8,16,balanced,0.09743616104125977
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,8,16,balanced,0.11743231773376465
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,8,16,balanced,0.13678976058959963
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,8,16,balanced,0.1823129653930664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,8,16,balanced,0.22013696670532226
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,8,16,balanced,0.3030284881591797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,8,16,balanced,0.06406911849975586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,8,16,balanced,0.3875904083251953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,8,16,power_law_1.01,0.05827583789825439
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,8,16,power_law_1.01,0.06338560104370117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,8,16,balanced,0.5753343963623048
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,8,16,power_law_1.01,0.06185728073120117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,8,16,power_law_1.01,0.06268671989440919
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,8,16,balanced,0.7374809265136719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,8,16,power_law_1.01,0.10353535652160643
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,8,16,power_law_1.01,0.1018560028076172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,8,16,power_law_1.01,0.11090432167053224
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,8,16,power_law_1.01,0.11172351837158204
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,8,16,power_law_1.01,0.11509759902954102
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,8,16,power_law_1.01,0.1192140769958496
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,8,16,power_law_1.01,0.12093695640563965
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,8,16,power_law_1.01,0.11766143798828126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,4,32,power_law_1.2,2.0821746826171874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,8,16,power_law_1.01,0.12273664474487304
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,8,16,power_law_1.01,0.1281382369995117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,8,16,power_law_1.01,0.13385984420776367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,8,16,power_law_1.01,0.14888704299926758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,8,16,power_law_1.01,0.20345087051391603
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,8,16,power_law_1.01,0.16434688568115235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,8,16,power_law_1.01,0.24941696166992186
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,8,16,power_law_1.01,0.10551039695739746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,8,16,power_law_1.01,0.3231052780151367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,8,16,power_law_1.01,0.04019199848175049
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,8,16,power_law_1.01,0.42901119232177737
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,8,16,power_law_1.01,0.5764096069335938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,8,16,power_law_1.01,0.04345856189727783
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,8,16,power_law_1.01,0.0414796781539917
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,8,16,power_law_1.01,0.8076236724853516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,4,32,power_law_1.2,1.113966064453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,8,16,power_law_1.01,0.04358399868011474
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,8,16,power_law_1.01,0.057608962059020996
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,8,16,power_law_1.01,0.06487296104431153
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,8,16,power_law_1.01,0.09740159988403321
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,8,16,power_law_1.01,0.06404863834381104
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,8,16,power_law_1.01,0.06681983947753907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,8,16,power_law_1.01,0.06837120056152343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,8,16,power_law_1.01,1.1948095703125001
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,8,16,power_law_1.01,0.07388800144195556
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,8,16,power_law_1.01,0.0763263988494873
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,8,16,power_law_1.01,0.0734438419342041
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,8,16,power_law_1.01,0.07411200046539307
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,8,16,power_law_1.01,0.07805312156677247
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,8,16,power_law_1.01,1.5203852844238281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,8,16,power_law_1.01,0.09740927696228027
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,8,16,power_law_1.01,0.11324543952941894
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,8,16,power_law_1.01,0.14157823562622068
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,8,16,power_law_1.01,0.17585535049438478
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,8,16,power_law_1.01,0.2384934425354004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,8,16,power_law_1.01,0.3030131149291992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,8,16,power_law_1.01,0.48283264160156253
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,8,16,power_law_1.01,0.06580095767974854
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,8,16,power_law_1.01,0.0713753604888916
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,8,16,power_law_1.2,0.058178558349609374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,8,16,power_law_1.01,0.5407206344604492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,8,16,power_law_1.2,0.06316415786743164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,8,16,power_law_1.2,0.060209918022155764
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,8,16,power_law_1.2,0.06411263942718506
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,8,16,power_law_1.01,0.08173439979553224
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,8,16,power_law_1.2,0.09879679679870605
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,8,16,power_law_1.2,0.10175999641418457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,8,16,power_law_1.01,0.8100531005859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,8,16,power_law_1.2,0.10299391746520996
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,8,16,power_law_1.2,0.10919551849365235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,8,16,power_law_1.2,0.11357312202453614
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,8,16,power_law_1.2,0.11532032012939453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,8,16,power_law_1.2,0.12120063781738281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,8,16,power_law_1.2,0.11911168098449706
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,8,16,power_law_1.2,0.1209216022491455
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,8,16,power_law_1.2,0.12532735824584962
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,8,16,power_law_1.2,0.13013631820678712
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,8,16,power_law_1.01,1.076099853515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,8,16,power_law_1.2,0.13598208427429198
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,8,16,power_law_1.2,0.15315584182739256
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,8,16,power_law_1.2,0.17522560119628908
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,8,16,power_law_1.2,0.22529279708862307
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,8,16,power_law_1.2,0.27732864379882816
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,8,16,power_law_1.2,0.4287411117553711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,8,16,power_law_1.2,0.10732031822204591
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,8,16,power_law_1.2,0.040357117652893064
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,8,16,power_law_1.2,0.5245708847045899
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,8,16,power_law_1.2,0.7750835418701172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,8,16,power_law_1.2,0.04279808044433594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,8,16,power_law_1.2,0.04332159996032715
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,8,16,power_law_1.2,0.059976959228515626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,8,16,power_law_1.2,0.06154111862182617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,8,16,power_law_1.2,0.06281216144561767
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,8,16,power_law_1.2,0.06495103836059571
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,8,16,power_law_1.2,0.06700672149658203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,8,16,power_law_1.2,0.06935423851013184
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,8,16,power_law_1.2,0.07276159763336182
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,8,16,power_law_1.2,0.07530111789703368
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,8,16,power_law_1.2,0.07340928077697753
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,8,16,power_law_1.2,0.0721676778793335
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,8,16,power_law_1.2,0.07642752170562744
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,8,16,power_law_1.2,0.07922688007354736
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,8,16,power_law_1.2,0.04419712066650391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,8,16,power_law_1.2,0.08600704193115234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,8,16,power_law_1.2,0.10110079765319824
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,8,16,power_law_1.2,0.12221440315246582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,8,16,power_law_1.2,0.15875328063964844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,8,16,power_law_1.2,0.20762624740600585
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,8,16,power_law_1.2,0.2806502342224121
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,8,16,power_law_1.2,0.37672702789306645
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,8,16,power_law_1.2,0.537916145324707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,16,8,balanced,0.02775808095932007
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,16,8,balanced,0.02905087947845459
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,8,16,power_law_1.2,0.7249075317382813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,16,8,balanced,0.03575551986694336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,16,8,balanced,0.05387135982513428
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,16,8,balanced,0.06562431812286376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,16,8,balanced,0.1064038372039795
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,16,8,balanced,0.11113344192504884
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,16,8,balanced,0.11596032142639161
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,16,8,balanced,0.118406400680542
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,16,8,balanced,0.12198911666870119
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,16,8,balanced,0.1147430419921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,16,8,balanced,0.11774335861206056
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,16,8,balanced,0.12045439720153808
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,16,8,balanced,0.12567168235778808
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,16,8,balanced,0.1293785572052002
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,16,8,balanced,0.136309757232666
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,16,8,balanced,0.1438924789428711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,16,8,balanced,0.16178943634033202
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,8,16,power_law_1.2,0.9675430297851563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,16,8,balanced,0.17966848373413086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,16,8,balanced,0.22314367294311524
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,16,8,balanced,0.25796735763549805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,16,8,balanced,0.33574527740478516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,16,8,balanced,0.421233901977539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,16,8,balanced,0.5788134384155273
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,16,8,balanced,0.02854144096374512
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,16,8,balanced,0.7915187072753906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,8,16,power_law_1.2,1.5671168518066405
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,16,8,balanced,0.028935680389404295
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,16,8,balanced,0.03056256055831909
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,16,8,balanced,0.03487871885299683
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,16,8,balanced,0.046786561012268066
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,16,8,balanced,0.0733516788482666
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,16,8,balanced,1.1457933044433593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,16,8,balanced,0.07515391826629639
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,16,8,balanced,0.07894527912139893
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,16,8,balanced,0.08040960311889649
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,16,8,balanced,0.07613952159881592
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,16,8,balanced,0.07889535903930664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,16,8,balanced,0.08321408271789552
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,16,8,balanced,0.08101247787475586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,16,8,balanced,1.4997158813476563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,16,8,balanced,0.0874176025390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,16,8,balanced,0.0925055980682373
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,16,8,balanced,0.10231167793273925
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,16,8,balanced,0.13982463836669923
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,16,8,balanced,0.17577856063842773
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,16,8,balanced,0.2119513511657715
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,16,8,balanced,0.300576000213623
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,16,8,balanced,0.372360954284668
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,16,8,balanced,0.5294604873657226
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,16,8,balanced,0.08572287559509277
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,16,8,power_law_1.01,0.05214208126068115
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,16,8,balanced,0.7207603454589844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,8,16,power_law_1.2,1.1458112335205077
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,16,8,power_law_1.01,0.05874815940856933
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,16,8,power_law_1.01,0.05763840198516845
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,16,8,balanced,1.0204659271240235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,16,8,power_law_1.01,0.06726528167724609
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,16,8,power_law_1.01,0.06376704216003418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,16,8,power_law_1.01,0.10405247688293458
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,16,8,power_law_1.01,0.10674816131591798
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,16,8,power_law_1.01,0.11144319534301758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,16,8,power_law_1.01,0.11531007766723633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,16,8,power_law_1.01,0.11853055953979492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,16,8,power_law_1.01,0.12096639633178712
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,16,8,balanced,1.3650265502929688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,16,8,power_law_1.01,0.11744768142700195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,16,8,power_law_1.01,0.11789823532104493
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,16,8,power_law_1.01,0.1218496036529541
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,16,8,balanced,0.1214361572265625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,16,8,power_law_1.01,0.12659711837768556
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,16,8,power_law_1.01,0.1364236831665039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,16,8,power_law_1.01,0.1413708782196045
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,16,8,power_law_1.01,0.15638527870178223
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,16,8,power_law_1.01,0.17644927978515626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,16,8,power_law_1.01,0.20824575424194336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,16,8,power_law_1.01,0.25177215576171874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,16,8,power_law_1.01,0.34918785095214844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,16,8,power_law_1.01,0.45491710662841794
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,8,16,power_law_1.2,2.1348889160156252
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,16,8,power_law_1.01,0.6040524673461914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,16,8,power_law_1.01,0.032940800189971926
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,16,8,power_law_1.01,0.04360832214355469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,16,8,power_law_1.01,0.8275827026367187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,16,8,power_law_1.01,0.040239357948303224
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,16,8,power_law_1.01,0.045561599731445315
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,16,8,power_law_1.01,0.05011072158813477
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,16,8,power_law_1.01,0.07375487804412842
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,16,8,power_law_1.01,0.07575295925140381
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,16,8,power_law_1.01,0.07645952224731445
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,16,8,power_law_1.01,1.1359693145751952
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,16,8,power_law_1.01,0.07919871807098389
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,16,8,power_law_1.01,0.08344832420349121
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,16,8,power_law_1.01,0.08082304000854493
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,16,8,power_law_1.01,0.0785152006149292
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,16,8,power_law_1.01,0.08270591735839844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,16,8,power_law_1.01,0.08412544250488281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,16,8,power_law_1.01,0.09249407768249511
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,16,8,power_law_1.01,0.10128000259399414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,16,8,power_law_1.01,1.7041651916503906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,16,8,power_law_1.01,0.13623295783996583
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,16,8,power_law_1.01,0.15908608436584473
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,16,8,power_law_1.01,0.19910144805908203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,16,8,power_law_1.01,0.07153535842895507
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,16,8,power_law_1.01,0.2907507133483887
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,16,8,power_law_1.01,0.3751359939575195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,8,16,power_law_1.2,1.5185626220703123
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,16,8,power_law_1.2,0.052227840423583985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,16,8,power_law_1.01,0.49964542388916017
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,16,8,power_law_1.2,0.05558784008026123
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,16,8,power_law_1.2,0.06373760223388672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,16,8,power_law_1.2,0.05880959987640381
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,16,8,power_law_1.01,0.692032012939453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,16,8,power_law_1.2,0.06580992221832276
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,16,8,power_law_1.2,0.1043673610687256
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,16,8,power_law_1.2,0.10844160079956056
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,16,8,power_law_1.2,0.11134079933166503
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,16,8,power_law_1.2,0.1187929630279541
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,16,8,power_law_1.01,0.9743078613281251
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,16,8,power_law_1.2,0.12186367988586426
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,16,8,power_law_1.2,0.11774592399597168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,16,8,power_law_1.2,0.12051456451416016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,16,8,power_law_1.2,0.12272255897521973
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,16,8,power_law_1.01,1.279784927368164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,16,8,power_law_1.2,0.12811264038085937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,16,8,power_law_1.2,0.13366527557373048
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,16,8,power_law_1.2,0.14271103858947753
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,16,8,power_law_1.2,0.16169599533081053
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,16,8,power_law_1.2,0.17716224670410158
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,16,8,power_law_1.2,0.26966272354125975
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,16,8,power_law_1.2,0.22695295333862306
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,16,8,power_law_1.2,0.3597004699707031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,16,8,power_law_1.01,0.1189516830444336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,16,8,power_law_1.2,0.4755699157714844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,16,8,power_law_1.2,0.03291520118713379
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,16,8,power_law_1.2,0.11298687934875488
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,16,8,power_law_1.2,0.03841792106628418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,16,8,power_law_1.2,0.6714521789550781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,16,8,power_law_1.2,0.0397273588180542
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,16,8,power_law_1.2,0.04442880153656006
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,16,8,power_law_1.2,0.04770688056945801
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,16,8,power_law_1.2,0.07227519989013673
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,16,8,power_law_1.2,0.07557248115539551
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,16,8,power_law_1.2,0.9424819183349609
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,16,8,power_law_1.2,0.07665152072906493
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,16,8,power_law_1.2,0.07932544231414795
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,16,8,power_law_1.2,0.08184191703796387
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,16,8,power_law_1.2,0.08134400367736816
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,16,8,power_law_1.2,0.08286463737487793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,16,8,power_law_1.2,0.08568703651428222
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,16,8,power_law_1.2,0.09050239562988281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,16,8,power_law_1.2,0.10171648025512696
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,16,8,power_law_1.2,0.11596927642822266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,16,8,power_law_1.2,0.14163071632385255
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,16,8,power_law_1.2,1.3701426696777343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,16,8,power_law_1.2,0.18839296340942382
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,16,8,power_law_1.2,0.22106367111206054
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,16,8,power_law_1.2,0.07167232036590576
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,16,8,power_law_1.2,0.3107289505004883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,16,8,power_law_1.2,0.4160383987426758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,32,4,balanced,0.0291865611076355
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,16,8,power_law_1.2,0.5704230499267579
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,32,4,balanced,0.03079040050506592
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,32,4,balanced,0.04196479797363281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,32,4,balanced,0.05841152191162109
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,32,4,balanced,0.08007424354553223
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,32,4,balanced,0.11569408416748046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,32,4,balanced,0.11838720321655274
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,32,4,balanced,0.12064512252807617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,16,8,power_law_1.2,0.7074444580078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,32,4,balanced,0.12349696159362793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,16,8,power_law_1.2,0.0816140842437744
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,32,4,balanced,0.11859968185424805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,32,4,balanced,0.12087167739868163
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,32,4,balanced,0.12370304107666015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,32,4,balanced,0.12927103996276856
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,32,4,balanced,0.13155200004577636
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,32,4,balanced,0.07068543910980225
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,32,4,balanced,0.1388326358795166
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,32,4,balanced,0.1461043167114258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,32,4,balanced,0.1637593650817871
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,32,4,balanced,0.18057216644287108
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,32,4,balanced,0.22770687103271486
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,32,4,balanced,0.2637990379333496
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,32,4,balanced,0.34468990325927734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,32,4,balanced,0.44987136840820313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,32,4,balanced,0.6319475173950195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,16,8,power_law_1.2,1.0976371002197265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,32,4,balanced,0.02976896047592163
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,32,4,balanced,0.8485158538818359
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,32,4,balanced,0.03136895895004273
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,32,4,balanced,0.034630401134490965
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,32,4,balanced,0.04085760116577149
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,32,4,balanced,0.057057280540466306
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,32,4,balanced,0.07354496002197267
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,32,4,balanced,0.0933568000793457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,32,4,balanced,0.09625344276428223
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,32,4,balanced,0.09098624229431153
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,32,4,balanced,0.09406080245971679
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,32,4,balanced,0.0960153579711914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,32,4,balanced,0.10037247657775879
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,32,4,balanced,1.5958335876464844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,32,4,balanced,0.10182784080505372
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,32,4,balanced,0.1070246410369873
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,32,4,balanced,0.11540736198425292
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,32,4,balanced,0.1348249626159668
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,32,4,balanced,0.15132287979125975
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,32,4,balanced,0.09152000427246094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,32,4,balanced,1.2408128356933594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,32,4,balanced,0.09784192085266114
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,32,4,balanced,0.3195264053344727
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,32,4,balanced,0.38874752044677735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,32,4,balanced,0.5550092697143555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,32,4,balanced,0.7558067321777344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,32,4,power_law_1.01,0.03229439973831177
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,32,4,balanced,0.18759424209594727
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,32,4,balanced,0.22300031661987302
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,32,4,power_law_1.01,0.05634560108184814
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,32,4,power_law_1.01,0.05503615856170654
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,32,4,balanced,1.061454086303711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,32,4,power_law_1.01,0.06185984134674072
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,32,4,power_law_1.01,0.06477439880371094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,32,4,power_law_1.01,0.11355135917663575
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,32,4,power_law_1.01,0.11954943656921387
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,32,4,power_law_1.01,0.12186240196228026
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,32,4,power_law_1.01,0.1225881576538086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,32,4,balanced,1.428970184326172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,16,8,power_law_1.2,1.8724838256835938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,32,4,power_law_1.01,0.12152064323425293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,32,4,power_law_1.01,0.12646656036376952
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,32,4,power_law_1.01,0.12932736396789551
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,32,4,power_law_1.01,0.14672256469726563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,32,4,power_law_1.01,0.16599807739257813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,32,4,power_law_1.01,0.18823936462402344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,32,4,power_law_1.01,0.22462335586547852
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,32,4,power_law_1.01,0.07688960075378418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,32,4,power_law_1.01,0.27140480041503906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,32,4,power_law_1.01,0.116428804397583
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,32,4,power_law_1.01,0.34256385803222655
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,32,4,power_law_1.01,0.11962752342224121
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,32,4,power_law_1.01,0.4586419296264649
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,32,4,power_law_1.01,0.033196799755096436
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,32,4,power_law_1.01,0.6740787506103516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,16,8,power_law_1.2,1.52840576171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,32,4,power_law_1.01,0.0417471981048584
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,32,4,power_law_1.01,0.8615679931640624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,32,4,power_law_1.01,0.04807424068450928
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,32,4,power_law_1.01,0.050448641777038575
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,32,4,power_law_1.01,0.061903362274169926
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,32,4,power_law_1.01,0.08737024307250976
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,32,4,power_law_1.01,0.09055999755859376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,32,4,power_law_1.01,0.09193216323852539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,32,4,power_law_1.01,0.09392767906188965
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,32,4,power_law_1.01,0.09922816276550292
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,32,4,power_law_1.01,0.13737215995788574
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,32,4,power_law_1.01,0.0906009578704834
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,32,4,power_law_1.01,0.09365504264831544
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,32,4,power_law_1.01,0.09727231979370117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,32,4,power_law_1.01,0.09940735816955566
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,32,4,power_law_1.01,0.10520319938659668
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,32,4,power_law_1.01,0.11629055976867675
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,32,4,power_law_1.01,1.6827929687499998
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,32,4,power_law_1.01,0.14034175872802734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,32,4,power_law_1.01,0.03847424030303955
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,32,4,power_law_1.01,0.19676288604736328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,32,4,power_law_1.01,0.24504959106445315
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,32,4,power_law_1.01,0.32297344207763673
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,32,4,power_law_1.01,0.403056640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,32,4,power_law_1.01,1.2678924560546876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,32,4,power_law_1.01,0.583365135192871
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,32,4,power_law_1.2,0.03593728065490723
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,32,4,power_law_1.01,0.752907485961914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,32,4,power_law_1.2,0.05330944061279297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,32,4,power_law_1.2,0.05640704154968261
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,32,4,power_law_1.2,0.060570878982543944
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,32,4,power_law_1.01,0.15501824378967285
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,32,4,power_law_1.2,0.06509312152862548
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,32,4,power_law_1.2,0.11338239669799806
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,32,4,power_law_1.2,0.11655936241149903
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,32,4,power_law_1.01,1.1492966461181642
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,32,4,power_law_1.2,0.11867520332336426
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,32,4,power_law_1.2,0.12127615928649901
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,32,4,power_law_1.2,0.1192191982269287
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,32,4,power_law_1.2,0.12094079971313476
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,32,4,power_law_1.2,0.1316710376739502
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,32,4,power_law_1.01,1.5186253356933594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,32,4,power_law_1.2,0.13837823867797852
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,32,4,power_law_1.2,0.1491110420227051
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,32,4,power_law_1.2,0.16994943618774414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,32,4,power_law_1.2,0.18978048324584962
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,32,4,power_law_1.2,0.23039615631103513
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,32,4,power_law_1.2,0.07728511810302734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,32,4,power_law_1.2,0.3658240127563476
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,32,4,power_law_1.2,0.12266240119934083
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,32,4,power_law_1.2,0.4632575988769531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,32,4,power_law_1.2,0.126112003326416
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,32,4,power_law_1.2,0.03343231916427612
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,32,4,power_law_1.2,0.03612799882888794
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,32,4,power_law_1.2,0.041648640632629394
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,32,4,power_law_1.2,0.9562931060791016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,32,4,power_law_1.2,0.045505280494689944
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,32,4,power_law_1.2,0.05090816020965576
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,32,4,power_law_1.2,0.2696281623840332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,32,4,power_law_1.2,0.060797438621521
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,32,4,power_law_1.2,0.08756480216979981
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,32,4,power_law_1.2,0.08877440452575683
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,32,4,power_law_1.2,0.09192831993103028
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,32,4,power_law_1.2,0.09586303710937501
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,32,4,power_law_1.2,1.317532196044922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,32,4,power_law_1.2,0.0924403190612793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,32,4,power_law_1.2,0.0909836769104004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,32,4,power_law_1.2,0.09187840461730958
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,32,4,power_law_1.2,0.10136447906494142
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,32,4,power_law_1.2,0.11822976112365721
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,32,4,power_law_1.2,0.10745984077453614
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,32,4,power_law_1.2,0.6799436950683594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,32,4,power_law_1.2,0.14333439826965333
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,32,4,power_law_1.2,0.16148351669311523
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,32,4,power_law_1.2,0.20459903717041014
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,32,4,power_law_1.2,0.24815488815307618
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,32,4,power_law_1.2,0.3318220901489258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,32,4,power_law_1.2,0.4023091125488281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,1,128,balanced,0.06755199909210205
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,1,128,balanced,0.06793727874755859
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,1,128,balanced,0.06742144107818604
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,1,128,balanced,0.06766975879669189
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,32,4,power_law_1.2,0.09664896011352539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,1,128,balanced,0.07819263935089112
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,1,128,balanced,0.09123583793640136
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,32,4,power_law_1.2,0.6267520141601562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,1,128,balanced,0.15017215728759764
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,1,128,balanced,0.14995967864990234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,1,128,balanced,0.1520038414001465
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,1,128,balanced,0.15325568199157716
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,1,128,balanced,0.15326335906982422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,1,128,balanced,0.06762752056121826
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,1,128,balanced,0.1568716812133789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,1,128,balanced,0.15638784408569334
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,1,128,balanced,0.1577894401550293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,1,128,balanced,0.16470144271850587
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,1,128,balanced,0.15084416389465333
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,1,128,balanced,0.16810367584228517
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,32,4,power_law_1.2,1.1627661132812501
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,1,128,balanced,0.1747635269165039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,1,128,balanced,0.18057472229003907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,1,128,balanced,0.18620800018310546
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,1,128,balanced,0.21776256561279297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,1,128,balanced,0.22441856384277345
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,1,128,balanced,0.2974310493469238
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,1,128,balanced,0.04001152038574219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,1,128,balanced,0.4001254272460938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,1,128,balanced,0.03964416027069092
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,1,128,balanced,0.0401036787033081
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,1,128,balanced,0.5115763092041016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,1,128,balanced,0.03982719898223877
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,1,128,balanced,0.040400638580322265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,1,128,balanced,0.04807551860809326
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,1,128,balanced,0.057058558464050294
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,1,128,balanced,0.06808576107025147
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,1,128,balanced,0.06822527885437012
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,1,128,balanced,0.06783743858337402
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,1,128,balanced,0.07046144008636475
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,1,128,balanced,0.06929024219512939
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,1,128,balanced,0.06971776008605957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,1,128,balanced,0.07104256153106689
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,1,128,balanced,0.07637631893157959
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,1,128,balanced,0.07298175811767578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,1,128,balanced,0.07778175830841064
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,32,4,power_law_1.2,0.789981460571289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,1,128,balanced,0.0818342399597168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,1,128,balanced,0.08889087677001953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,1,128,balanced,0.08574080467224121
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,1,128,balanced,0.09394047737121583
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,1,128,balanced,0.15744000434875488
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,1,128,balanced,0.11725824356079102
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,1,128,balanced,0.13992064476013183
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,1,128,balanced,0.16756479263305663
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,1,128,balanced,0.23845632553100585
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,1,128,power_law_1.01,0.09155200004577638
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,1,128,power_law_1.01,0.09139840126037597
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,1,128,balanced,0.07443583965301513
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,1,128,power_law_1.01,0.09183103561401366
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,1,128,power_law_1.01,0.14675583839416503
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,1,128,power_law_1.01,0.15003647804260253
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,1,128,balanced,0.10032511711120606
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,1,128,power_law_1.01,0.15523072242736818
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,1,128,power_law_1.01,0.15323391914367676
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,1,128,power_law_1.01,0.15600128173828126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,1,128,power_law_1.01,0.15645952224731446
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,1,128,power_law_1.01,0.16125951766967775
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,1,128,power_law_1.01,0.15661696434020994
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,1,128,power_law_1.01,0.15744383811950685
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,1,128,power_law_1.01,0.16564224243164063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,1,128,power_law_1.01,0.16788223266601562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,1,128,power_law_1.01,0.0913152027130127
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,1,128,power_law_1.01,0.17046016693115235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,1,128,power_law_1.01,0.17672063827514647
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,32,4,power_law_1.2,1.753912353515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,1,128,power_law_1.01,0.20197631835937502
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,1,128,power_law_1.01,0.2569932746887207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,1,128,power_law_1.01,0.3168063926696777
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,1,128,power_law_1.01,0.4728128051757812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,1,128,power_law_1.01,0.5373311996459961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,1,128,power_law_1.01,0.05689983844757081
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,1,128,power_law_1.01,0.7481049346923828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,1,128,power_law_1.01,0.056396799087524416
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,1,128,power_law_1.01,0.9895629119873046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,1,128,power_law_1.01,0.1867532730102539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,1,128,power_law_1.01,0.05610239982604981
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,1,128,power_law_1.01,0.06660607814788819
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,1,128,power_law_1.01,0.06854911804199218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,1,128,power_law_1.01,0.06834943771362305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,32,4,power_law_1.2,1.563941192626953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,1,128,power_law_1.01,0.07144703865051269
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,1,128,power_law_1.01,1.5565478515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,1,128,power_law_1.01,0.074170880317688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,1,128,power_law_1.01,0.0730508804321289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,1,128,power_law_1.01,0.07910016059875488
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,1,128,power_law_1.01,0.07419136047363281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,1,128,power_law_1.01,0.08310655593872071
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,1,128,power_law_1.01,1.9066099548339843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,1,128,power_law_1.01,0.08794112205505371
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,1,128,power_law_1.01,0.056930561065673825
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,1,128,power_law_1.01,0.09340543746948242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,1,128,power_law_1.01,0.09990528106689453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,1,128,power_law_1.01,0.12120320320129394
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,1,128,power_law_1.01,0.14794367790222168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,1,128,power_law_1.01,0.2238540840148926
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,1,128,power_law_1.01,0.2558822441101074
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,1,128,power_law_1.01,0.06910079956054688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,1,128,power_law_1.01,0.40921089172363284
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,1,128,power_law_1.01,0.43634815216064454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,1,128,power_law_1.2,0.09183103561401366
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,1,128,power_law_1.2,0.09124480247497559
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,1,128,power_law_1.01,0.07654399871826172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,1,128,power_law_1.2,0.09138048171997071
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,1,128,power_law_1.01,0.6660160064697266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,1,128,power_law_1.2,0.09155584335327148
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,1,128,power_law_1.2,0.14456959724426272
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,1,128,power_law_1.01,0.09236224174499512
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,1,128,power_law_1.2,0.14007040023803713
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,1,128,power_law_1.2,0.15280896186828613
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,1,128,power_law_1.2,0.1556544017791748
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,1,128,power_law_1.01,0.8997042846679687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,1,128,power_law_1.2,0.15560192108154297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,1,128,power_law_1.2,0.15819135665893555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,1,128,power_law_1.2,0.1607206344604492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,1,128,power_law_1.2,0.15739904403686525
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,1,128,power_law_1.2,0.16908031463623047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,1,128,power_law_1.2,0.16990976333618163
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,1,128,power_law_1.2,0.1809881591796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,1,128,power_law_1.2,0.17779455184936524
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,1,128,power_law_1.2,0.2398080062866211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,1,128,power_law_1.2,0.3252479934692383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,1,128,power_law_1.2,0.41337726593017576
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,1,128,power_law_1.2,0.5528665542602539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,1,128,power_law_1.2,0.677702407836914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,1,128,power_law_1.2,0.1567321586608887
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,1,128,power_law_1.2,0.05637119770050049
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,1,128,power_law_1.2,0.05540224075317383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,1,128,power_law_1.2,0.056029438972473145
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,1,128,power_law_1.2,1.1283865356445313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,1,128,power_law_1.2,0.19548288345336914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,1,128,power_law_1.2,0.06622591972351075
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,1,128,power_law_1.2,0.06772607803344727
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,1,128,power_law_1.2,0.0688640022277832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,1,128,power_law_1.2,1.59146240234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,1,128,power_law_1.2,0.0702400016784668
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,1,128,power_law_1.2,0.07159423828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,1,128,power_law_1.2,0.0740825605392456
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,1,128,power_law_1.2,0.0762943983078003
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,1,128,power_law_1.2,0.07765503883361816
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,1,128,power_law_1.2,0.08369536399841308
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,1,128,power_law_1.2,0.05690239906311036
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,1,128,power_law_1.2,2.281144256591797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,1,128,power_law_1.2,0.08714879989624023
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,1,128,power_law_1.2,0.09057536125183105
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,1,128,power_law_1.2,0.09591808319091796
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,1,128,power_law_1.2,0.06699776172637939
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,1,128,power_law_1.2,0.11163776397705079
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,1,128,power_law_1.2,0.1462169647216797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,1,128,power_law_1.2,0.1876518440246582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,1,128,power_law_1.2,0.25059200286865235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,1,128,power_law_1.2,0.33961856842041016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,1,128,power_law_1.2,0.4969523239135742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,2,64,balanced,0.05652991771697998
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,1,128,power_law_1.2,0.08948863983154297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,2,64,balanced,0.055770878791809086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,1,128,power_law_1.2,0.7475328063964843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,2,64,balanced,0.0555072021484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,2,64,balanced,0.06336383819580078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,2,64,balanced,0.07529856204986572
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,1,128,power_law_1.2,1.011962890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,2,64,balanced,0.14192511558532714
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,2,64,balanced,0.14165887832641602
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,2,64,balanced,0.14111231803894042
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,2,64,balanced,0.14376192092895507
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,2,64,balanced,0.1465766429901123
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,1,128,power_law_1.2,4.4025946044921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,2,64,balanced,0.1460326385498047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,2,64,balanced,0.15311488151550293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,2,64,balanced,0.15701248168945312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,2,64,balanced,0.1568998432159424
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,1,128,power_law_1.2,1.6537791442871093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,2,64,balanced,0.15976832389831544
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,2,64,balanced,0.17070335388183594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,2,64,balanced,0.17987712860107422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,2,64,balanced,0.19839231491088866
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,2,64,balanced,0.0912499237060547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,2,64,balanced,0.24713600158691404
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,2,64,balanced,0.3203865432739258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,2,64,balanced,0.03227519989013672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,2,64,balanced,0.03230720043182373
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,2,64,balanced,0.4522905731201172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,2,64,balanced,0.14982912063598633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,2,64,balanced,0.03222656011581421
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,2,64,balanced,0.5801587295532227
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,2,64,balanced,0.032052481174468996
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,2,64,balanced,0.038821120262146
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,2,64,balanced,0.16284927368164062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,2,64,balanced,0.056698880195617675
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,2,64,balanced,0.05998079776763916
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,2,64,balanced,0.07235328197479249
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,2,64,balanced,0.07242879867553711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,2,64,balanced,0.07520512104034424
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,2,64,balanced,0.0741107177734375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,2,64,balanced,0.07503488063812255
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,2,64,balanced,0.07647615909576416
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,2,64,balanced,0.07593472003936767
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,2,64,balanced,0.0771340799331665
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,2,64,balanced,0.23353343963623047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,2,64,balanced,0.08167167663574218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,2,64,balanced,0.0855948829650879
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,2,64,balanced,0.08179327964782715
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,2,64,balanced,0.08716927528381348
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,2,64,balanced,0.09251456260681153
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,2,64,balanced,0.10624896049499513
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,2,64,balanced,0.12008447647094726
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,2,64,balanced,0.1454412841796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,2,64,balanced,0.1820044708251953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,2,64,balanced,0.22811647415161135
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,2,64,power_law_1.01,0.09042431831359862
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,2,64,balanced,0.3200384140014648
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,2,64,power_law_1.01,0.09061632156372071
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,2,64,power_law_1.01,0.09189248085021973
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,2,64,power_law_1.01,0.1331392002105713
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,2,64,power_law_1.01,0.14010111808776854
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,2,64,power_law_1.01,0.13498751640319823
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,2,64,power_law_1.01,0.14011136054992676
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,2,64,balanced,0.07798783779144287
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,2,64,power_law_1.01,0.1426252841949463
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,2,64,power_law_1.01,0.14598400115966798
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,2,64,power_law_1.01,0.1505254364013672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,2,64,power_law_1.01,0.15235839843750001
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,2,64,power_law_1.01,0.15771519660949707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,2,64,power_law_1.01,0.16042240142822267
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,2,64,power_law_1.01,0.1637081527709961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,2,64,power_law_1.01,0.16238464355468749
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,2,64,power_law_1.01,0.18011392593383788
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,2,64,power_law_1.01,0.08978560447692871
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,2,64,power_law_1.01,0.1987660789489746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,2,64,power_law_1.01,0.22856191635131834
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,2,64,balanced,0.04812928199768066
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,2,64,power_law_1.01,0.36560127258300784
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,2,64,power_law_1.01,0.45720958709716797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,2,64,power_law_1.01,0.14117888450622557
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,2,64,power_law_1.01,0.05812607765197754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,2,64,power_law_1.01,0.6335193634033203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,2,64,power_law_1.01,0.7837670135498047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,2,64,power_law_1.01,0.05856512069702149
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,2,64,power_law_1.01,0.05874815940856933
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,2,64,power_law_1.01,0.05931647777557373
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,2,64,power_law_1.01,0.06044928073883057
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,2,64,power_law_1.01,1.1356915283203126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,2,64,power_law_1.01,0.27047679901123045
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,2,64,power_law_1.01,0.06986495971679688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,2,64,power_law_1.01,0.07170944213867188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,2,64,power_law_1.01,0.07602431774139404
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,2,64,power_law_1.01,0.07405695915222169
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,2,64,power_law_1.01,0.07570687770843507
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,2,64,power_law_1.01,0.07535103797912598
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,2,64,power_law_1.01,0.07686912059783936
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,2,64,power_law_1.01,0.08223744392395019
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,2,64,power_law_1.01,0.0842086410522461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,2,64,power_law_1.01,0.0840601634979248
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,2,64,power_law_1.01,1.8120089721679686
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,2,64,power_law_1.01,0.08589568138122558
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,2,64,power_law_1.01,0.09259519577026368
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,2,64,power_law_1.01,0.09800704002380371
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,2,64,power_law_1.01,0.1228313636779785
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,2,64,power_law_1.01,0.14175871849060057
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,2,64,power_law_1.01,0.1843187141418457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,2,64,power_law_1.01,0.2279257583618164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,2,64,power_law_1.01,0.3472447967529297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,2,64,power_law_1.2,0.09079551696777344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,2,64,power_law_1.2,0.09079680442810059
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,2,64,power_law_1.01,0.46508800506591796
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,2,64,power_law_1.2,0.09125632286071778
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,2,64,power_law_1.01,0.6155891036987304
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,2,64,power_law_1.2,0.09136768341064452
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,2,64,power_law_1.2,0.135032320022583
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,2,64,power_law_1.2,0.13215744018554687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,2,64,power_law_1.2,0.1273356819152832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,2,64,power_law_1.01,0.9166028594970703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,2,64,power_law_1.2,0.13629568099975586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,2,64,power_law_1.2,0.1412070369720459
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,2,64,power_law_1.2,0.14653056144714355
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,2,64,power_law_1.2,0.15576576232910155
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,2,64,power_law_1.01,0.06691840171813965
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,2,64,power_law_1.2,0.15545087814331054
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,2,64,power_law_1.2,0.15048831939697266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,2,64,power_law_1.2,0.16289663314819336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,2,64,power_law_1.2,0.16265472412109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,2,64,power_law_1.2,0.16912128448486327
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,2,64,power_law_1.2,0.19144960403442385
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,2,64,power_law_1.2,0.21672960281372072
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,2,64,power_law_1.2,0.26723968505859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,2,64,power_law_1.2,0.3161651229858399
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,2,64,power_law_1.2,0.42753025054931637
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,2,64,power_law_1.2,0.058652157783508296
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,2,64,power_law_1.2,0.05834752082824707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,2,64,power_law_1.2,0.1381888008117676
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,2,64,power_law_1.2,0.7988454437255859
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,2,64,power_law_1.2,0.05835391998291015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,2,64,power_law_1.2,0.05940991878509522
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,2,64,power_law_1.2,0.06878848075866699
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,2,64,power_law_1.2,1.205523223876953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,2,64,power_law_1.2,0.06739327907562256
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,2,64,power_law_1.2,0.06991744041442871
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,2,64,power_law_1.2,0.07099135875701904
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,2,64,power_law_1.2,0.07155327796936035
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,2,64,power_law_1.2,0.07677696228027345
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,2,64,power_law_1.2,0.07513343811035156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,2,64,power_law_1.2,0.07842048168182372
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,2,64,power_law_1.2,0.07800064086914063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,2,64,power_law_1.2,0.5933337783813476
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,2,64,power_law_1.2,0.08097663879394532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,2,64,power_law_1.2,0.08536704063415528
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,2,64,power_law_1.2,0.08332544326782226
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,2,64,power_law_1.2,0.09879167556762695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,2,64,power_law_1.2,0.1045145606994629
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,2,64,power_law_1.2,0.13522560119628907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,2,64,power_law_1.2,0.17821311950683594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,2,64,power_law_1.2,0.2502604866027832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,2,64,power_law_1.2,0.32306049346923826
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,2,64,power_law_1.2,0.4547647857666016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,4,32,balanced,0.03110527992248535
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,2,64,power_law_1.2,2.702080078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,2,64,power_law_1.2,1.598849334716797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,4,32,balanced,0.03101696014404297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,4,32,balanced,0.03129215955734253
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,4,32,balanced,0.058607358932495114
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,4,32,balanced,0.0776255989074707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,4,32,balanced,0.09436287879943847
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,2,64,power_law_1.2,0.08577152252197265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,4,32,balanced,0.15057024002075195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,4,32,balanced,0.15034624099731447
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,4,32,balanced,0.1547225570678711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,2,64,power_law_1.2,0.9937728118896484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,4,32,balanced,0.15988479614257814
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,4,32,balanced,0.15912320137023925
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,4,32,balanced,0.16265344619750977
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,4,32,balanced,0.16389888763427735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,4,32,balanced,0.16679040908813475
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,4,32,balanced,0.1630963134765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,4,32,balanced,0.1677516746520996
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,4,32,balanced,0.17516799926757814
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,2,64,power_law_1.2,0.5504844665527344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,4,32,balanced,0.18744447708129883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,4,32,balanced,0.20145280838012694
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,4,32,balanced,0.2243622398376465
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,4,32,balanced,0.06221183776855469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,4,32,balanced,0.2624831962585449
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,4,32,balanced,0.29762176513671873
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,4,32,balanced,0.15038208007812498
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,4,32,balanced,0.3971968078613281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,4,32,balanced,0.03080575942993164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,4,32,balanced,0.031252479553222655
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,4,32,balanced,0.03089024066925049
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,4,32,balanced,0.5665497589111328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,4,32,balanced,0.03875328063964843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,4,32,balanced,0.051429119110107425
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,2,64,power_law_1.2,1.1999705505371092
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,4,32,balanced,0.06197120189666748
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,4,32,balanced,0.08144639968872071
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,4,32,balanced,0.07678336143493653
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,4,32,balanced,0.07689856052398682
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,4,32,balanced,0.7395059204101562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,4,32,balanced,0.0802995204925537
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,4,32,balanced,0.07810815811157226
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,4,32,balanced,0.08015359878540039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,4,32,balanced,0.08400896072387695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,4,32,balanced,0.08670720100402832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,4,32,balanced,0.08975359916687012
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,4,32,balanced,0.08380928039550781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,4,32,balanced,0.08873600006103516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,4,32,balanced,0.09313023567199707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,4,32,balanced,0.10469504356384278
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,4,32,balanced,0.11968128204345703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,4,32,balanced,0.140250883102417
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,4,32,balanced,0.1658252716064453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,4,32,balanced,0.2075302314758301
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,4,32,balanced,0.264898567199707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,4,32,power_law_1.01,0.07834368228912354
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,4,32,balanced,0.3496755218505859
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,4,32,balanced,0.03185791969299316
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,4,32,power_law_1.01,0.0921996784210205
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,4,32,balanced,0.4916159820556641
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,4,32,power_law_1.01,0.09370623588562012
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,4,32,power_law_1.01,0.09397120475769043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,4,32,power_law_1.01,0.14071935653686524
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,4,32,power_law_1.01,0.14695296287536622
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,4,32,power_law_1.01,0.14525823593139647
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,4,32,power_law_1.01,0.14446463584899902
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,4,32,power_law_1.01,0.146560001373291
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,4,32,power_law_1.01,0.14687487602233887
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,4,32,power_law_1.01,0.15255167961120605
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,4,32,power_law_1.01,0.15557120323181153
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,4,32,power_law_1.01,0.1600614356994629
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,4,32,power_law_1.01,0.16003456115722656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,4,32,power_law_1.01,0.16425983428955077
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,4,32,power_law_1.01,0.09078016281127929
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,4,32,power_law_1.01,0.17335296630859376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,4,32,power_law_1.01,0.18411264419555665
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,4,32,power_law_1.01,0.25290111541748045
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,4,32,power_law_1.01,0.30942848205566403
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,4,32,power_law_1.01,0.14424448013305663
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,4,32,power_law_1.01,0.40990974426269533
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,4,32,power_law_1.01,0.05036159992218018
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,4,32,power_law_1.01,0.5548953628540039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,4,32,power_law_1.01,0.05964799880981445
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,4,32,power_law_1.01,0.060275201797485356
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,4,32,power_law_1.01,0.7387136077880859
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,4,32,power_law_1.01,0.062067198753356936
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,4,32,power_law_1.01,0.07325183868408203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,4,32,power_law_1.01,1.0097510528564453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,4,32,power_law_1.01,0.20864128112792968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,4,32,power_law_1.01,0.07669888019561769
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,4,32,power_law_1.01,0.07659264087677002
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,4,32,power_law_1.01,0.07893375873565674
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,4,32,power_law_1.01,0.07698815822601318
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,4,32,power_law_1.01,0.08293375968933106
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,4,32,power_law_1.01,0.08896256446838378
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,4,32,power_law_1.01,1.5174118041992188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,4,32,power_law_1.01,0.08283264160156249
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,4,32,power_law_1.01,0.08375807762145995
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,4,32,power_law_1.01,0.08842240333557129
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,4,32,power_law_1.01,0.060906238555908196
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,4,32,power_law_1.01,0.10480383872985839
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,4,32,power_law_1.01,0.13741567611694336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,4,32,power_law_1.01,0.14649215698242188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,4,32,power_law_1.01,0.20777471542358397
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,4,32,power_law_1.01,0.07438335895538331
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,4,32,power_law_1.01,0.27171072006225583
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,4,32,power_law_1.01,0.07587711811065674
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,4,32,power_law_1.01,0.3912960052490234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,4,32,power_law_1.2,0.078951678276062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,4,32,power_law_1.01,0.4407603073120117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,4,32,power_law_1.2,0.09094143867492675
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,4,32,power_law_1.2,0.09145855903625488
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,4,32,power_law_1.2,0.09384448051452636
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,4,32,power_law_1.01,0.09735039710998536
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,4,32,power_law_1.01,0.7886975860595703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,4,32,power_law_1.2,0.1436582374572754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,4,32,power_law_1.2,0.13270015716552735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,4,32,power_law_1.2,0.14690431594848632
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,4,32,power_law_1.2,0.1328985595703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,4,32,power_law_1.01,1.0141363525390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,4,32,power_law_1.2,0.14127488136291505
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,4,32,power_law_1.2,0.150383358001709
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,4,32,power_law_1.2,0.14719743728637696
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,4,32,power_law_1.2,0.1586854362487793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,4,32,power_law_1.2,0.16156288146972658
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,4,32,power_law_1.2,0.1624064064025879
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,4,32,power_law_1.2,0.16690687179565428
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,4,32,power_law_1.2,0.19242624282836912
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,4,32,power_law_1.2,0.09217408180236816
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,4,32,power_law_1.2,0.2190553665161133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,4,32,power_law_1.2,0.2801228713989258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,4,32,power_law_1.2,0.4027443313598633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,4,32,power_law_1.2,0.46073345184326175
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,4,32,power_law_1.2,0.14772992134094237
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,4,32,power_law_1.2,0.04980480194091797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,4,32,power_law_1.2,0.6842739105224609
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,4,32,power_law_1.2,0.05961088180541992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,4,32,power_law_1.2,0.17940223693847657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,4,32,power_law_1.2,0.9212992095947266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,4,32,power_law_1.2,0.06023808002471924
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,4,32,power_law_1.2,0.06093311786651612
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,4,32,power_law_1.2,0.0767948818206787
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,4,32,power_law_1.2,0.06956031799316406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,4,32,power_law_1.2,0.07489408016204833
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,4,32,power_law_1.2,0.07240320205688476
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,4,32,power_law_1.2,0.0761190414428711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,4,32,power_law_1.2,0.0776793622970581
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,4,32,power_law_1.2,1.522677764892578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,4,32,power_law_1.2,0.08019840240478515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,4,32,power_law_1.2,0.08397055625915527
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,4,32,power_law_1.2,0.08806015968322753
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,4,32,power_law_1.2,0.08468223571777343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,4,32,power_law_1.2,0.09032832145690918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,4,32,power_law_1.2,0.08498687744140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,4,32,power_law_1.2,0.10410112380981444
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,4,32,power_law_1.2,1.7727449035644534
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,4,32,power_law_1.2,0.11981951713562011
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,4,32,power_law_1.2,0.13928832054138182
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,4,32,power_law_1.2,0.17150592803955078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,4,32,power_law_1.2,0.23515647888183594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,8,16,balanced,0.03078399896621704
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,8,16,balanced,0.030945279598236085
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,4,32,power_law_1.2,0.07494783878326416
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,4,32,power_law_1.2,0.43527168273925787
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,8,16,balanced,0.03191807985305786
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,8,16,balanced,0.055381760597229004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,8,16,balanced,0.0648307180404663
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,4,32,power_law_1.2,0.5952985763549805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,8,16,balanced,0.08109824180603027
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,8,16,balanced,0.09763584136962891
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,8,16,balanced,0.14982912063598633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,8,16,balanced,0.1512985610961914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,8,16,balanced,0.1569395160675049
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,8,16,balanced,0.15120767593383788
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,8,16,balanced,0.15993599891662597
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,8,16,balanced,0.16274944305419922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,8,16,balanced,0.1553151988983154
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,4,32,power_law_1.2,0.9745394897460937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,8,16,balanced,0.15944448471069336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,8,16,balanced,0.16057727813720704
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,8,16,balanced,0.1677132797241211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,8,16,balanced,0.17971967697143554
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,8,16,balanced,0.1886028861999512
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,4,32,power_law_1.2,1.146118392944336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,8,16,balanced,0.2078668785095215
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,8,16,balanced,0.23260671615600587
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,4,32,power_law_1.2,0.3515238571166992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,8,16,balanced,0.32933502197265624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,8,16,balanced,0.03103872060775757
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,8,16,balanced,0.4104204940795898
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,8,16,balanced,0.0313702392578125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,8,16,balanced,0.034296319484710694
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,8,16,balanced,0.548139533996582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,8,16,balanced,0.0430131196975708
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,8,16,balanced,0.057227520942687994
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,8,16,balanced,0.07165440082550048
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,8,16,balanced,0.08974592208862305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,8,16,balanced,0.09082624435424805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,8,16,balanced,0.8173542022705078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,8,16,balanced,0.09621760368347168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,8,16,balanced,0.09954560279846192
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,8,16,balanced,0.09305215835571289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,8,16,balanced,1.058275833129883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,8,16,balanced,0.09616512298583985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,8,16,balanced,0.09790847778320313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,8,16,balanced,0.27510528564453124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,8,16,balanced,0.11102975845336913
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,8,16,balanced,0.14418944358825683
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,8,16,balanced,0.030933759212493896
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,8,16,balanced,0.16757631301879883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,8,16,balanced,0.20700031280517578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,8,16,balanced,0.24967935562133786
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,8,16,balanced,0.32866943359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,8,16,balanced,0.09024255752563477
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,8,16,balanced,0.43673984527587895
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,8,16,balanced,0.09843456268310546
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,8,16,power_law_1.01,0.06429823875427246
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,8,16,power_law_1.01,0.07627264022827149
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,8,16,balanced,0.597070083618164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,8,16,power_law_1.01,0.08950400352478027
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,8,16,balanced,0.10056703567504884
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,8,16,power_law_1.01,0.09092351913452149
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,8,16,balanced,0.12395392417907716
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,8,16,power_law_1.01,0.094900484085083
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,8,16,balanced,0.8234931182861327
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,8,16,power_law_1.01,0.09272831916809082
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,8,16,power_law_1.01,0.13003647804260254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,8,16,power_law_1.01,0.13742079734802246
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,8,16,power_law_1.01,0.141780481338501
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,8,16,power_law_1.01,0.15141375541687013
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,8,16,power_law_1.01,0.1506175994873047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,8,16,power_law_1.01,0.15169919967651366
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,8,16,power_law_1.01,0.15133312225341797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,8,16,power_law_1.01,0.15629440307617187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,8,16,power_law_1.01,0.1648409652709961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,8,16,power_law_1.01,0.17909120559692382
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,8,16,power_law_1.01,0.1922547149658203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,8,16,power_law_1.01,0.22481920242309572
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,8,16,power_law_1.01,0.2565555191040039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,8,16,power_law_1.01,0.33071231842041016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,8,16,power_law_1.01,0.42581630706787105
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,8,16,power_law_1.01,0.13780608177185058
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,8,16,power_law_1.01,0.14543231964111328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,8,16,power_law_1.01,0.6242790222167969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,8,16,power_law_1.01,0.05412864208221435
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,8,16,power_law_1.01,0.7787289428710937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,8,16,power_law_1.01,0.06454271793365478
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,8,16,power_law_1.01,0.06402431964874268
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,8,16,power_law_1.01,0.06830592155456543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,8,16,power_law_1.01,0.06711167812347411
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,8,16,power_law_1.01,1.0815001678466796
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,8,16,power_law_1.01,0.0830016040802002
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,8,16,power_law_1.01,0.08346624374389648
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,8,16,power_law_1.01,0.08528127670288085
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,8,16,power_law_1.01,1.3976666259765627
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,8,16,power_law_1.01,0.09281408309936523
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,8,16,power_law_1.01,0.04100607872009278
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,8,16,power_law_1.01,0.09303168296813966
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,8,16,power_law_1.01,0.09491583824157715
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,8,16,power_law_1.01,0.10277631759643555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,8,16,power_law_1.01,0.12078592300415039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,8,16,power_law_1.01,0.12652416229248048
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,8,16,power_law_1.01,0.15741056442260742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,8,16,power_law_1.01,0.19111040115356445
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,8,16,power_law_1.01,0.0836736011505127
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,8,16,power_law_1.01,0.25132415771484373
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,8,16,power_law_1.01,0.3035417556762695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,8,16,power_law_1.01,0.08773504257202148
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,8,16,power_law_1.01,0.4141913604736328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,8,16,power_law_1.2,0.0641868782043457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,8,16,power_law_1.01,0.0938419246673584
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,8,16,power_law_1.2,0.07669760227203369
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,8,16,power_law_1.2,0.08808320045471192
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,8,16,power_law_1.2,0.09011584281921386
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,8,16,power_law_1.01,0.8044544219970703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,8,16,power_law_1.2,0.09859328269958496
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,8,16,power_law_1.2,0.1430476760864258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,8,16,power_law_1.2,0.1342899227142334
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,8,16,power_law_1.01,1.047704315185547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,8,16,power_law_1.2,0.1384819221496582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,8,16,power_law_1.2,0.1367155170440674
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,8,16,power_law_1.2,0.13975808143615723
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,8,16,power_law_1.2,0.14534784317016602
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,8,16,power_law_1.01,0.09234047889709472
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,8,16,power_law_1.2,0.14901887893676757
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,8,16,power_law_1.2,0.1540428829193115
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,8,16,power_law_1.2,0.1467123222351074
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,8,16,power_law_1.2,0.1551257610321045
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,8,16,power_law_1.2,0.15934847831726073
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,8,16,power_law_1.2,0.1658291244506836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,8,16,power_law_1.2,0.19037311553955077
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,8,16,power_law_1.2,0.19240447998046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,8,16,power_law_1.2,0.2428223991394043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,8,16,power_law_1.2,0.26494592666625977
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,8,16,power_law_1.2,0.35130241394042966
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,8,16,power_law_1.2,0.48246910095214846
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,8,16,power_law_1.2,0.04127232074737549
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,8,16,power_law_1.2,0.05418240070343018
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,8,16,power_law_1.2,0.6878374481201173
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,8,16,power_law_1.2,0.06441215991973877
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,8,16,power_law_1.2,0.06367616176605224
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,8,16,power_law_1.2,0.8830528259277344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,8,16,power_law_1.2,0.06798208236694336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,8,16,power_law_1.2,0.08697728157043458
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,8,16,power_law_1.01,0.5753702545166016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,8,16,power_law_1.2,0.08387711524963379
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,8,16,power_law_1.2,0.08200703620910645
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,8,16,power_law_1.2,1.2448358154296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,8,16,power_law_1.2,0.08153471946716309
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,8,16,power_law_1.2,0.0845580768585205
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,8,16,power_law_1.2,0.09001983642578125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,8,16,power_law_1.2,0.09207679748535155
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,8,16,power_law_1.2,0.09008383750915527
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,8,16,power_law_1.2,0.0946009635925293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,8,16,power_law_1.2,0.100316162109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,8,16,power_law_1.2,0.12728192329406737
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,8,16,power_law_1.2,0.1366988754272461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,8,16,power_law_1.2,0.17709440231323242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,8,16,power_law_1.2,1.8207936096191406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,8,16,power_law_1.2,0.21755392074584962
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,8,16,power_law_1.2,0.2671180725097656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,8,16,power_law_1.2,0.346824951171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,8,16,power_law_1.2,0.5108646392822266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,16,8,balanced,0.03202431917190552
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,8,16,power_law_1.2,0.09511808395385743
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,16,8,balanced,0.04003583908081055
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,16,8,balanced,0.032956159114837645
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,8,16,power_law_1.2,0.09844863891601563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,16,8,balanced,0.06455167770385742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,16,8,balanced,0.07246592044830322
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,16,8,balanced,0.08872320175170897
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,16,8,balanced,0.10824192047119141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,16,8,balanced,0.1540595245361328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,16,8,balanced,0.1572761631011963
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,16,8,balanced,0.15934720039367675
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,16,8,balanced,0.15416192054748534
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,16,8,balanced,0.1574118423461914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,16,8,balanced,0.16177280426025392
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,16,8,balanced,0.16862335205078124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,16,8,balanced,0.17290367126464842
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,8,16,power_law_1.2,1.27085693359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,16,8,balanced,0.17791744232177736
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,16,8,balanced,0.18712320327758789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,16,8,balanced,0.20609535217285155
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,16,8,balanced,0.22526336669921876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,8,16,power_law_1.2,0.6455372619628906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,16,8,balanced,0.30804351806640623
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,16,8,balanced,0.3911539077758789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,16,8,balanced,0.4890803146362305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,8,16,power_law_1.2,0.9600422668457032
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,16,8,balanced,0.03211519956588745
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,16,8,balanced,0.6492889404296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,16,8,balanced,0.03250047922134399
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,16,8,balanced,0.03548543930053711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,16,8,balanced,0.042007040977478025
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,16,8,balanced,0.05200255870819091
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,16,8,balanced,0.8665280151367188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,16,8,balanced,0.07162367820739746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,16,8,balanced,0.0894694423675537
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,16,8,balanced,0.11414015769958497
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,16,8,balanced,0.26613248825073244
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,16,8,balanced,0.11591296195983887
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,16,8,balanced,0.11337471961975099
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,16,8,balanced,0.11835904121398926
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,16,8,balanced,1.321026611328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,16,8,balanced,0.12338047981262208
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,16,8,balanced,0.12798720359802246
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,16,8,balanced,0.1320307159423828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,16,8,balanced,0.1418278408050537
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,16,8,balanced,0.16044799804687498
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,16,8,balanced,0.1811622428894043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,16,8,balanced,0.22228607177734375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,16,8,balanced,0.2616153526306152
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,16,8,balanced,1.7195059204101562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,16,8,balanced,0.34023040771484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,16,8,balanced,0.4200051116943359
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,16,8,power_law_1.01,0.05579520225524902
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,16,8,balanced,0.5793011093139648
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,16,8,power_law_1.01,0.06610047817230225
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,16,8,balanced,0.779183349609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,16,8,power_law_1.01,0.0803059196472168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,16,8,power_law_1.01,0.07777791976928711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,16,8,power_law_1.01,0.09026432037353516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,16,8,power_law_1.01,0.09946623802185059
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,16,8,balanced,1.0970240020751953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,16,8,power_law_1.01,0.1475660800933838
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,16,8,power_law_1.01,0.1420198440551758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,16,8,power_law_1.01,0.14486016273498534
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,16,8,power_law_1.01,0.14466431617736816
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,16,8,balanced,1.49267333984375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,16,8,power_law_1.01,0.14269184112548827
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,16,8,balanced,0.11752575874328612
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,16,8,power_law_1.01,0.1539046382904053
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,16,8,power_law_1.01,0.1586969566345215
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,16,8,power_law_1.01,0.16176000595092774
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,16,8,power_law_1.01,0.1689574432373047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,16,8,power_law_1.01,0.19162879943847658
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,16,8,power_law_1.01,0.20651008605957033
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,16,8,power_law_1.01,0.2418726348876953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,16,8,power_law_1.01,0.09360383987426758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,16,8,power_law_1.01,0.27922431945800785
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,16,8,power_law_1.01,0.35136512756347654
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,16,8,power_law_1.01,0.46537727355957037
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,16,8,power_law_1.01,0.14840191841125489
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,16,8,power_law_1.01,0.03661184072494507
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,16,8,balanced,0.11538944244384766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,16,8,power_law_1.01,0.6304051208496093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,16,8,power_law_1.01,0.04842239856719971
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,16,8,power_law_1.01,0.06481664180755616
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,16,8,power_law_1.01,0.8189401245117187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,16,8,power_law_1.01,0.07167359828948974
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,16,8,power_law_1.01,0.07866112232208253
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,16,8,power_law_1.01,0.10949631690979005
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,16,8,power_law_1.01,0.10604160308837891
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,16,8,power_law_1.01,0.10282112121582032
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,16,8,power_law_1.01,0.10735615730285644
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,16,8,power_law_1.01,0.10827136039733887
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,16,8,power_law_1.01,0.10668160438537597
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,16,8,power_law_1.01,0.1139673614501953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,16,8,power_law_1.01,1.49625732421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,16,8,power_law_1.01,0.1179532814025879
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,16,8,power_law_1.01,0.11696255683898926
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,16,8,power_law_1.01,0.06265344142913817
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,16,8,power_law_1.01,0.1289292812347412
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,16,8,power_law_1.01,0.15868800163269042
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,16,8,power_law_1.01,0.13902976036071776
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,16,8,power_law_1.01,0.19595392227172853
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,16,8,power_law_1.01,0.07744383811950684
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,16,8,power_law_1.01,0.22196224212646481
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,16,8,power_law_1.01,1.192579803466797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,16,8,power_law_1.01,0.31063295364379884
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,16,8,power_law_1.01,0.3946521759033203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,16,8,power_law_1.2,0.060046081542968754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,16,8,power_law_1.01,0.5418841552734375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,16,8,power_law_1.2,0.06582655906677246
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,16,8,power_law_1.2,0.07285888195037842
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,16,8,power_law_1.01,0.7047398376464844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,16,8,power_law_1.2,0.07883391857147218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,16,8,power_law_1.2,0.09508992195129394
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,16,8,power_law_1.2,0.10204416275024415
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,16,8,power_law_1.01,0.9935897827148438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,16,8,power_law_1.2,0.14926464080810548
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,16,8,power_law_1.2,0.1390105628967285
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,16,8,power_law_1.2,0.14278400421142579
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,16,8,power_law_1.2,0.14738944053649902
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,16,8,power_law_1.2,0.14799615859985352
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,16,8,power_law_1.2,0.1447500801086426
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,16,8,power_law_1.2,0.1531481647491455
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,16,8,power_law_1.01,1.3569369506835938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,16,8,power_law_1.2,0.1603750419616699
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,16,8,power_law_1.2,0.15944191932678223
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,16,8,power_law_1.2,0.17280639648437499
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,16,8,power_law_1.2,0.1935372734069824
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,16,8,power_law_1.2,0.2108582305908203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,16,8,power_law_1.2,0.2794867134094238
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,16,8,power_law_1.2,0.26025728225708006
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,16,8,power_law_1.2,0.37161342620849613
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,16,8,power_law_1.2,0.47595905303955083
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,16,8,power_law_1.2,0.03676671981811523
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,16,8,power_law_1.2,0.6837145233154297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,16,8,power_law_1.2,0.04809216022491455
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,16,8,power_law_1.2,0.05801216125488281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,16,8,power_law_1.2,0.060962557792663574
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,16,8,power_law_1.2,0.8834099578857423
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,16,8,power_law_1.2,0.07008255958557129
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,16,8,power_law_1.2,0.07303936004638671
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,16,8,power_law_1.2,0.08291839599609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,16,8,power_law_1.2,0.10893695831298829
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,16,8,power_law_1.2,0.09083264350891114
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,16,8,power_law_1.2,0.09935999870300292
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,16,8,power_law_1.2,0.10611583709716796
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,16,8,power_law_1.2,0.10757887840270997
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,16,8,power_law_1.2,0.10687616348266601
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,16,8,power_law_1.2,1.3165682983398437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,16,8,power_law_1.2,0.11043071746826172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,16,8,power_law_1.2,0.11194623947143553
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,16,8,power_law_1.2,0.11805952072143555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,16,8,power_law_1.2,0.11796607971191406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,16,8,power_law_1.2,0.12364928245544433
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,16,8,power_law_1.2,0.14568703651428222
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,16,8,power_law_1.2,1.743089904785156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,16,8,power_law_1.2,0.1638912010192871
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,16,8,power_law_1.2,0.21527551651000976
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,16,8,power_law_1.2,0.25373952865600585
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,16,8,power_law_1.2,0.30454015731811523
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,32,4,balanced,0.030202879905700687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,32,4,balanced,0.032264959812164304
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,16,8,power_law_1.2,0.39710720062255855
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,32,4,balanced,0.045722880363464356
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,32,4,balanced,0.06400256156921387
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,32,4,balanced,0.07907455921173095
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,16,8,power_law_1.2,0.5804479980468751
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,32,4,balanced,0.09725055694580079
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,32,4,balanced,0.11850111961364747
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,32,4,balanced,0.11979264259338378
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,32,4,balanced,0.12115712165832519
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,16,8,power_law_1.2,0.7481126403808593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,32,4,balanced,0.16354944229125976
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,32,4,balanced,0.16619264602661132
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,32,4,balanced,0.16972415924072265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,32,4,balanced,0.1737868881225586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,32,4,balanced,0.18316543579101563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,32,4,balanced,0.1914841651916504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,32,4,balanced,0.2101568031311035
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,32,4,balanced,0.22956672668457032
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,16,8,power_law_1.2,1.0924224090576171
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,32,4,balanced,0.26908800125122073
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,32,4,balanced,0.3089062309265137
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,16,8,power_law_1.2,1.3653273010253906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,32,4,balanced,0.39115009307861326
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,32,4,balanced,0.030392320156097413
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,32,4,balanced,0.0318937611579895
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,32,4,balanced,0.03585792064666748
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,32,4,balanced,0.17839359283447265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,32,4,balanced,0.8562918090820313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,32,4,balanced,0.04523776054382324
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,32,4,balanced,0.0633139181137085
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,32,4,balanced,0.09105279922485351
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,32,4,balanced,0.11741439819335937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,32,4,balanced,0.11849856376647949
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,32,4,balanced,1.3110208129882812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,32,4,balanced,0.11989248275756836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,32,4,balanced,0.12122495651245117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,32,4,balanced,0.492097282409668
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,32,4,balanced,0.16249216079711915
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,32,4,balanced,0.16024831771850584
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,32,4,balanced,0.12311936378479005
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,32,4,balanced,0.16603776931762695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,32,4,balanced,0.17023359298706056
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,32,4,balanced,1.7935845947265627
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,32,4,balanced,0.1740403175354004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,32,4,balanced,0.17699840545654297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,32,4,balanced,0.18624000549316405
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,32,4,balanced,0.20481536865234373
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,32,4,balanced,0.2250534439086914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,32,4,balanced,0.264192008972168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,32,4,balanced,0.2995199966430664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,32,4,balanced,0.37624961853027344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,32,4,balanced,0.45160961151123047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,32,4,power_law_1.01,0.0331660795211792
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,32,4,balanced,0.6049062347412109
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,32,4,power_law_1.01,0.057123842239379885
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,32,4,balanced,0.6524838256835938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,32,4,power_law_1.01,0.06925055980682374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,32,4,balanced,0.8791500854492188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,32,4,power_law_1.01,0.07265920162200927
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,32,4,power_law_1.01,0.08615039825439454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,32,4,power_law_1.01,0.09833215713500977
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,32,4,power_law_1.01,0.09919360160827637
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,32,4,power_law_1.01,0.10876928329467775
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,32,4,balanced,1.187311325073242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,32,4,power_law_1.01,0.11109503746032715
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,32,4,power_law_1.01,0.14997119903564454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,32,4,power_law_1.01,0.15561087608337404
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,32,4,power_law_1.01,0.15733375549316406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,32,4,power_law_1.01,0.16455039978027344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,32,4,power_law_1.01,0.16577791213989257
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,32,4,balanced,1.6124748229980468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,32,4,power_law_1.01,0.17423744201660157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,32,4,power_law_1.01,0.17934335708618163
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,32,4,power_law_1.01,0.2028851127624512
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,32,4,power_law_1.01,0.22100864410400392
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,32,4,power_law_1.01,0.2660121536254883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,32,4,power_law_1.01,0.30652799606323244
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,32,4,power_law_1.01,0.3962470245361328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,32,4,power_law_1.01,0.10996095657348633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,32,4,power_law_1.01,0.49139839172363275
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,32,4,power_law_1.01,0.034666240215301514
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,32,4,power_law_1.01,0.041301760673522946
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,32,4,power_law_1.01,0.6784729766845703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,32,4,power_law_1.01,0.05674752235412598
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,32,4,power_law_1.01,0.06212480068206787
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,32,4,power_law_1.01,0.07504127979278565
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,32,4,power_law_1.01,0.08712575912475587
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,32,4,power_law_1.01,0.8779122924804688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,32,4,power_law_1.01,0.09276288032531739
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,32,4,power_law_1.01,0.10382847785949707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,32,4,power_law_1.01,0.10557184219360352
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,32,4,power_law_1.01,0.11079551696777343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,32,4,power_law_1.01,0.13952768325805665
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,32,4,power_law_1.01,0.14620287895202638
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,32,4,power_law_1.01,0.15020544052124024
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,32,4,power_law_1.01,0.15487232208251953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,32,4,power_law_1.01,0.1485043239593506
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,32,4,power_law_1.01,0.16274303436279297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,32,4,power_law_1.01,1.3631692504882813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,32,4,power_law_1.01,0.16898048400878907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,32,4,power_law_1.01,0.18843904495239258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,32,4,power_law_1.01,0.21523456573486327
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,32,4,power_law_1.01,1.6945114135742188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,32,4,power_law_1.01,0.25636863708496094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,32,4,power_law_1.01,0.30053119659423827
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,32,4,power_law_1.01,0.3908031845092773
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,32,4,power_law_1.2,0.033164799213409424
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,32,4,power_law_1.2,0.05762432098388672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,32,4,power_law_1.01,0.5163033676147462
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,32,4,power_law_1.2,0.06497536182403565
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,32,4,power_law_1.01,0.7059903717041015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,32,4,power_law_1.2,0.0846399974822998
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,32,4,power_law_1.2,0.08988927841186524
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,32,4,power_law_1.01,0.8735180664062501
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,32,4,power_law_1.2,0.09510527610778809
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,32,4,power_law_1.2,0.10714752197265624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,32,4,power_law_1.2,0.10714752197265624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,32,4,power_law_1.2,0.11086208343505859
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,32,4,power_law_1.2,0.15741824150085448
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,32,4,power_law_1.2,0.14845439910888672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,32,4,power_law_1.2,0.15804415702819824
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,32,4,power_law_1.2,0.16239999771118163
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,32,4,power_law_1.01,1.24063232421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,32,4,power_law_1.2,0.17067007064819334
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,32,4,power_law_1.2,0.17470592498779297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,32,4,power_law_1.2,0.18169599533081054
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,32,4,power_law_1.2,0.2041279983520508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,32,4,power_law_1.01,1.6508709716796877
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,32,4,power_law_1.2,0.22296960830688475
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,32,4,power_law_1.2,0.07322368144989014
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,32,4,power_law_1.2,0.2740313529968262
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,32,4,power_law_1.2,0.3080217552185059
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,32,4,power_law_1.2,0.4019916915893555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,32,4,power_law_1.2,0.034923520088195804
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,32,4,power_law_1.2,0.4888857650756836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,32,4,power_law_1.2,0.04095615863800049
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,32,4,power_law_1.2,0.7347110748291016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,32,4,power_law_1.2,0.050597119331359866
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,32,4,power_law_1.2,0.061177601814270025
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,32,4,power_law_1.2,0.9246771240234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,32,4,power_law_1.2,0.07456128120422363
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,32,4,power_law_1.2,0.0906278419494629
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,32,4,power_law_1.2,0.09009407997131348
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,32,4,power_law_1.2,0.102478084564209
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,32,4,power_law_1.2,0.10121855735778809
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,32,4,power_law_1.2,0.10836864471435548
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,32,4,power_law_1.2,1.3466073608398437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,32,4,power_law_1.2,0.13771264076232909
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,32,4,power_law_1.2,0.14988160133361816
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,32,4,power_law_1.2,0.1531289577484131
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,32,4,power_law_1.2,0.15343615531921387
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,32,4,power_law_1.2,0.16467199325561524
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,32,4,power_law_1.2,0.1938380813598633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,32,4,power_law_1.2,0.21879167556762696
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,32,4,power_law_1.2,0.2636953544616699
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,32,4,power_law_1.2,0.3108441543579102
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,32,4,power_law_1.2,0.4338022232055664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,32,4,power_law_1.2,0.15041791915893554
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,32,4,power_law_1.2,0.5304064178466797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,1,128,balanced,0.023223040103912355
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,32,4,power_law_1.2,0.7106150054931641
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,1,128,balanced,0.022842879295349124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,1,128,balanced,0.023170559406280516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,1,128,balanced,0.023133440017700197
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,1,128,balanced,0.04159743785858154
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,32,4,power_law_1.2,0.89728515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,1,128,balanced,0.0681382417678833
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,1,128,balanced,0.06880512237548828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,1,128,balanced,0.06865536212921143
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,1,128,balanced,0.06892928123474121
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,1,128,balanced,0.0698367977142334
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,1,128,balanced,0.0696614408493042
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,1,128,balanced,0.0697267198562622
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,1,128,balanced,0.07121664047241211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,1,128,balanced,0.07262080192565919
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,32,4,power_law_1.2,1.2781938934326171
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,1,128,balanced,0.07344128131866455
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,1,128,balanced,0.07492735862731933
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,1,128,balanced,0.07604224205017089
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,32,4,power_law_1.2,1.8075762939453124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,1,128,balanced,0.07924352169036866
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,1,128,balanced,0.08326656341552734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,1,128,balanced,0.022999041080474854
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,1,128,balanced,0.10087679862976073
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,32,4,power_law_1.2,1.7236160278320312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,1,128,balanced,0.10824192047119141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,1,128,balanced,0.1505356788635254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,1,128,balanced,0.025067520141601563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,1,128,balanced,0.02496000051498413
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,1,128,balanced,0.0248473596572876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,1,128,balanced,0.02460288047790527
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,1,128,balanced,0.025055999755859377
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,1,128,balanced,0.025228800773620608
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,1,128,balanced,0.03945215940475464
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,1,128,balanced,0.03911551952362061
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,1,128,balanced,0.039705600738525394
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,1,128,balanced,0.039682559967041016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,1,128,balanced,0.03984384059906006
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,1,128,balanced,0.08428544044494629
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,1,128,balanced,0.04007552146911621
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,1,128,balanced,0.0401804780960083
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,1,128,balanced,0.04078080177307129
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,1,128,balanced,0.04133632183074951
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,1,128,balanced,0.08820863723754882
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,1,128,balanced,0.041507840156555176
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,1,128,balanced,0.04320000171661377
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,1,128,balanced,0.04247039794921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,1,128,balanced,0.044710397720336914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,1,128,balanced,0.052828159332275394
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,1,128,balanced,0.05466112136840821
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,1,128,balanced,0.1881727981567383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,1,128,balanced,0.06591616153717042
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,1,128,balanced,0.07393152236938476
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,1,128,balanced,0.09146623611450196
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,1,128,balanced,0.11343999862670899
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,1,128,power_law_1.01,0.041402878761291506
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,1,128,power_law_1.01,0.041359357833862305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,1,128,power_law_1.01,0.041429758071899414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,1,128,power_law_1.01,0.04165376186370849
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,1,128,power_law_1.01,0.06908160209655761
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,32,4,power_law_1.2,0.1723148727416992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,1,128,power_law_1.01,0.06921088218688964
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,1,128,power_law_1.01,0.06969727993011474
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,1,128,power_law_1.01,0.07080448150634766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,1,128,power_law_1.01,0.07094912052154541
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,1,128,power_law_1.01,0.07430272102355957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,1,128,power_law_1.01,0.07372288227081299
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,1,128,power_law_1.01,0.07514368057250977
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,1,128,balanced,0.04763264179229736
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,1,128,power_law_1.01,0.07585919857025146
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,1,128,power_law_1.01,0.07933184146881103
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,1,128,power_law_1.01,0.08249343872070312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,1,128,power_law_1.01,0.08485504150390624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,1,128,power_law_1.01,0.08294272422790527
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,1,128,power_law_1.01,0.08636672019958495
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,1,128,power_law_1.01,0.09155584335327148
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,1,128,power_law_1.01,0.10440064430236817
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,1,128,power_law_1.01,0.12895999908447267
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,1,128,power_law_1.01,0.1787494468688965
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,1,128,power_law_1.01,0.2270182418823242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,1,128,power_law_1.01,0.3109836769104004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,1,128,power_law_1.01,0.024591360092163086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,1,128,power_law_1.01,0.43590656280517576
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,1,128,power_law_1.01,0.025075199604034426
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,1,128,power_law_1.01,0.6198374557495117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,1,128,power_law_1.01,0.025166079998016355
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,1,128,power_law_1.01,0.025315840244293213
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,1,128,balanced,0.0587775993347168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,1,128,power_law_1.01,0.03968512058258057
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,1,128,power_law_1.01,0.8355583953857423
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,1,128,power_law_1.01,0.03994368076324463
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,1,128,power_law_1.01,0.04039423942565918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,1,128,power_law_1.01,0.0406822395324707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,1,128,power_law_1.01,0.042700800895690914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,1,128,power_law_1.01,0.042629117965698245
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,1,128,power_law_1.01,0.04373631954193115
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,1,128,power_law_1.01,0.041818881034851076
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,1,128,power_law_1.01,0.044893441200256345
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,1,128,balanced,0.0804646396636963
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,1,128,power_law_1.01,0.054176001548767096
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,1,128,power_law_1.01,0.053084158897399904
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,1,128,power_law_1.01,0.05595903873443604
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,1,128,power_law_1.01,0.06929664134979248
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,1,128,power_law_1.01,0.07963776111602783
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,1,128,power_law_1.01,0.11049216270446778
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,1,128,power_law_1.01,0.13160063743591308
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,1,128,power_law_1.01,0.1926976013183594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,1,128,power_law_1.01,0.03982719898223877
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,1,128,power_law_1.01,0.29228160858154295
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,1,128,power_law_1.2,0.04129663944244385
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,1,128,power_law_1.01,0.4335295867919922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,1,128,power_law_1.2,0.04137087821960449
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,1,128,power_law_1.2,0.04067967891693115
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,1,128,power_law_1.01,0.04915584087371826
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,1,128,power_law_1.2,0.06830207824707031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,1,128,power_law_1.01,0.061459197998046874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,1,128,power_law_1.2,0.06863999843597413
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,1,128,power_law_1.2,0.06760191917419434
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,1,128,power_law_1.2,0.06984960079193116
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,1,128,power_law_1.2,0.07105663776397705
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,1,128,power_law_1.2,0.07350143909454346
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,1,128,power_law_1.2,0.07585792064666748
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,1,128,power_law_1.2,0.07870207786560059
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,1,128,power_law_1.2,0.08184320449829101
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,1,128,power_law_1.2,0.08549375534057617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,1,128,power_law_1.2,0.08600447654724122
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,1,128,power_law_1.2,0.0836518383026123
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,1,128,power_law_1.2,0.09372544288635254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,1,128,power_law_1.2,0.10532095909118652
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,1,128,power_law_1.2,0.1283827209472656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,1,128,power_law_1.2,0.16667776107788085
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,1,128,power_law_1.01,0.49849857330322267
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,1,128,power_law_1.2,0.3332735824584961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,1,128,power_law_1.2,0.49077121734619145
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,1,128,power_law_1.2,0.07236608028411864
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,1,128,power_law_1.2,0.02489727973937988
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,1,128,power_law_1.2,0.07406976222991943
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,1,128,power_law_1.2,0.02528127908706665
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,1,128,power_law_1.2,0.03907327890396119
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,1,128,power_law_1.2,0.03966464042663574
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,1,128,power_law_1.2,0.0410265588760376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,1,128,power_law_1.2,0.03983488082885742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,1,128,power_law_1.2,0.2605606460571289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,1,128,power_law_1.2,0.04068352222442627
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,1,128,power_law_1.2,0.04196608066558838
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,1,128,power_law_1.2,0.04338943958282471
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,1,128,power_law_1.2,0.04273663997650147
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,1,128,power_law_1.2,0.04410367965698242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,1,128,power_law_1.2,0.04815616130828858
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,1,128,power_law_1.2,1.6896319580078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,1,128,power_law_1.2,0.05257855892181397
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,1,128,power_law_1.2,0.7009471893310547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,1,128,power_law_1.2,0.05120639801025391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,1,128,power_law_1.2,0.05272831916809082
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,1,128,power_law_1.2,0.06644480228424073
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,1,128,power_law_1.2,0.0594316816329956
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,1,128,power_law_1.2,0.07683072090148926
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,1,128,power_law_1.2,0.09839872360229493
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,1,128,power_law_1.2,0.13997183799743654
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,1,128,power_law_1.2,0.19586687088012694
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,1,128,power_law_1.2,1.2966079711914062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,1,128,power_law_1.2,0.29052415847778323
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,1,128,power_law_1.2,0.04179840087890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,2,64,balanced,0.02082175970077515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,2,64,balanced,0.020887041091918947
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,2,64,balanced,0.020787200927734374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,1,128,power_law_1.2,0.46696830749511714
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,2,64,balanced,0.021607680320739748
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,2,64,balanced,0.03832832098007202
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,2,64,balanced,0.060862717628479
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,2,64,balanced,0.06112895965576172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,2,64,balanced,0.06200831890106201
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,1,128,power_law_1.2,0.02500864028930664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,2,64,balanced,0.06218880176544189
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,2,64,balanced,0.0629375982284546
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,2,64,balanced,0.06388607978820801
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,2,64,balanced,0.06439680099487305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,2,64,balanced,0.06733952045440675
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,2,64,balanced,0.06890495777130126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,2,64,balanced,0.07195519924163818
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,2,64,balanced,0.0758681583404541
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,2,64,balanced,0.07020031929016113
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,2,64,balanced,0.0746611213684082
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,2,64,balanced,0.0778495979309082
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,2,64,balanced,0.0859878444671631
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,2,64,balanced,0.09039999961853026
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,2,64,balanced,0.11176575660705565
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,2,64,balanced,0.020616960525512696
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,1,128,power_law_1.2,0.7330854034423828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,2,64,balanced,0.17844095230102539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,2,64,balanced,0.022906880378723144
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,2,64,balanced,0.022882559299468995
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,2,64,balanced,0.022783999443054197
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,2,64,balanced,0.02313215970993042
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,2,64,balanced,0.02376960039138794
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,2,64,balanced,0.023695359230041506
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,2,64,balanced,0.03627392053604126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,2,64,balanced,0.036831998825073244
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,1,128,power_law_1.2,0.9317414093017579
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,2,64,balanced,0.03666048049926758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,2,64,balanced,0.03674240112304687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,2,64,balanced,0.03858304023742676
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,2,64,balanced,0.03773184061050415
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,2,64,balanced,0.03881472110748291
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,2,64,balanced,0.03884543895721436
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,2,64,balanced,0.039521279335021975
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,2,64,balanced,0.04090367794036866
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,2,64,balanced,0.12493696212768554
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,2,64,balanced,0.042699518203735354
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,2,64,balanced,0.04783616065979004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,2,64,balanced,0.05146751880645752
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,2,64,balanced,0.22177919387817382
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,2,64,balanced,0.05771903991699219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,2,64,balanced,0.06387839794158937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,2,64,balanced,0.08051456451416016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,2,64,balanced,0.09533056259155273
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,2,64,balanced,0.06579967975616455
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,2,64,balanced,0.15443584442138672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,2,64,power_law_1.01,0.03765376091003418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,2,64,power_law_1.01,0.03837055921554565
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,2,64,power_law_1.01,0.03808383941650391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,2,64,balanced,0.037857279777526856
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,2,64,power_law_1.01,0.03834367990493774
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,2,64,power_law_1.01,0.05515776157379151
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,2,64,power_law_1.01,0.05595647811889648
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,2,64,power_law_1.01,0.060441598892211915
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,2,64,power_law_1.01,0.06245632171630859
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,2,64,balanced,0.04646783828735351
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,2,64,power_law_1.01,0.06422272205352783
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,2,64,power_law_1.01,0.06611711978912353
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,2,64,power_law_1.01,0.06833024024963379
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,2,64,power_law_1.01,0.07098239898681641
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,2,64,power_law_1.01,0.07465983867645264
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,2,64,power_law_1.01,0.07318528175354004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,2,64,power_law_1.01,0.07455743789672852
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,2,64,power_law_1.01,0.07423999786376953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,2,64,power_law_1.01,0.08047616004943847
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,2,64,power_law_1.01,0.08750975608825684
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,2,64,power_law_1.01,0.10834815979003906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,2,64,power_law_1.01,0.11830783843994142
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,2,64,balanced,0.12632320404052735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,2,64,power_law_1.01,0.1618713569641113
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,2,64,power_law_1.01,0.1973632049560547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,2,64,power_law_1.01,0.2887795257568359
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,2,64,power_law_1.01,0.023557119369506836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,2,64,power_law_1.01,0.023831040859222413
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,2,64,power_law_1.01,0.37549182891845706
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,2,64,power_law_1.01,0.023989760875701906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,2,64,power_law_1.01,0.023713281154632566
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,2,64,power_law_1.01,0.03611648082733154
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,2,64,power_law_1.01,0.5813068771362305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,2,64,power_law_1.01,0.06300672054290771
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,2,64,power_law_1.01,0.03670016050338745
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,2,64,power_law_1.01,0.03671295881271362
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,2,64,power_law_1.01,0.037740800380706784
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,2,64,power_law_1.01,0.734286117553711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,2,64,power_law_1.01,0.03863424062728882
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,2,64,power_law_1.01,0.0398528003692627
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,2,64,power_law_1.01,0.04104320049285889
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,2,64,power_law_1.01,0.04665855884552002
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,2,64,power_law_1.01,0.048325119018554685
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,2,64,power_law_1.01,0.0478656005859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,2,64,power_law_1.01,0.053393921852111816
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,2,64,power_law_1.01,0.06157567977905274
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,2,64,power_law_1.01,0.07465472221374511
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,2,64,power_law_1.01,0.08284671783447266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,2,64,power_law_1.01,0.1171072006225586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,2,64,power_law_1.01,0.22557439804077148
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,2,64,power_law_1.01,0.2773593521118164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,2,64,power_law_1.2,0.03805056095123291
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,2,64,power_law_1.01,0.3807872009277344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,2,64,power_law_1.01,0.03799936056137085
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,2,64,power_law_1.2,0.038064639568328854
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,2,64,power_law_1.01,0.03873791933059693
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,2,64,power_law_1.2,0.038176000118255615
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,2,64,power_law_1.2,0.057658882141113275
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,2,64,power_law_1.2,0.05134079933166504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,2,64,power_law_1.2,0.05802239894866943
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,2,64,power_law_1.2,0.06198272228240966
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,2,64,power_law_1.2,0.06386559963226318
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,2,64,power_law_1.01,0.13136384010314942
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,2,64,power_law_1.2,0.06205056190490723
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,2,64,power_law_1.2,0.06502143859863281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,2,64,power_law_1.2,0.06638720035552978
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,2,64,power_law_1.2,0.06965760231018067
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,2,64,power_law_1.2,0.07311744213104247
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,2,64,power_law_1.2,0.07613952159881592
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,2,64,power_law_1.2,0.07413119792938232
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,2,64,power_law_1.2,0.07299327850341797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,2,64,power_law_1.2,0.07644032001495361
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,2,64,power_law_1.2,0.08453887939453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,2,64,power_law_1.2,0.09383296012878418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,2,64,power_law_1.2,0.11710975646972657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,2,64,power_law_1.01,0.4992550277709961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,2,64,power_law_1.2,0.15573375701904296
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,2,64,power_law_1.01,0.04798079967498779
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,2,64,power_law_1.2,0.2826675224304199
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,2,64,power_law_1.2,0.023613440990447997
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,2,64,power_law_1.2,0.3805132675170898
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,2,64,power_law_1.2,0.47643009185791013
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,2,64,power_law_1.2,0.02367743968963623
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,2,64,power_law_1.2,0.023592960834503175
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,2,64,power_law_1.2,0.03565567970275879
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,2,64,power_law_1.2,0.03625727891921997
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,2,64,power_law_1.2,0.036839680671691896
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,2,64,power_law_1.2,0.037710080146789546
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,2,64,power_law_1.2,0.039248640537261965
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,2,64,power_law_1.2,0.19762432098388671
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,2,64,power_law_1.2,0.04033535957336426
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,2,64,power_law_1.2,1.1447795104980467
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,2,64,power_law_1.2,0.04052864074707031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,2,64,power_law_1.2,0.04494463920593262
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,2,64,power_law_1.2,0.049388799667358395
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,2,64,power_law_1.2,0.05023231983184815
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,2,64,power_law_1.2,0.04748672008514404
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,2,64,power_law_1.2,0.05090432167053223
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,2,64,power_law_1.2,0.060888319015502936
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,2,64,power_law_1.2,0.06876416206359863
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,2,64,power_law_1.2,0.08239359855651855
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,2,64,power_law_1.2,0.10237183570861816
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,2,64,power_law_1.2,0.1410764789581299
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,2,64,power_law_1.2,0.037103359699249265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,2,64,power_law_1.2,0.17144960403442383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,2,64,power_law_1.2,0.2950054359436035
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,4,32,balanced,0.019376640319824216
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,4,32,balanced,0.019528959989547727
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,2,64,power_law_1.2,0.038680319786071775
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,2,64,power_law_1.2,0.3798886489868164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,4,32,balanced,0.020650238990783693
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,4,32,balanced,0.022403841018676755
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,4,32,balanced,0.03761280059814453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,4,32,balanced,0.05884287834167481
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,4,32,balanced,0.06011903762817383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,4,32,balanced,0.06042751789093017
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,2,64,power_law_1.2,0.6248511886596679
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,4,32,balanced,0.06268928050994874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,4,32,balanced,0.061896958351135255
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,4,32,balanced,0.06274432182312012
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,4,32,balanced,0.06731391906738281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,4,32,balanced,0.0703001594543457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,4,32,balanced,0.0725107192993164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,4,32,balanced,0.06553215980529785
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,4,32,balanced,0.06961791992187501
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,4,32,balanced,0.0736959981918335
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,4,32,balanced,0.0783462381362915
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,4,32,balanced,0.0840396785736084
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,4,32,balanced,0.10033151626586914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,4,32,balanced,0.11132927894592286
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,2,64,power_law_1.2,0.9264025878906249
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,4,32,balanced,0.16699392318725587
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,4,32,balanced,0.019370239973068238
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,4,32,balanced,0.23861120223999022
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,4,32,balanced,0.022177278995513916
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,4,32,balanced,0.021934080123901366
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,4,32,balanced,0.3015372848510742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,4,32,balanced,0.057966079711914066
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,4,32,balanced,0.022196478843688965
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,4,32,balanced,0.022204160690307617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,4,32,balanced,0.022691841125488284
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,4,32,balanced,0.02467839956283569
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,4,32,balanced,0.03546367883682251
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,4,32,balanced,0.03668992042541504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,4,32,balanced,0.03833856105804444
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,4,32,balanced,0.03701632022857666
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,4,32,balanced,0.038096640110015866
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,4,32,balanced,0.04473599910736084
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,4,32,balanced,0.047146239280700684
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,4,32,balanced,0.046821122169494626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,4,32,balanced,0.04089344024658203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,4,32,balanced,0.04608640193939209
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,4,32,balanced,0.049247999191284184
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,4,32,balanced,0.05446911811828613
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,4,32,balanced,0.05848447799682617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,4,32,balanced,0.07372416019439697
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,4,32,balanced,0.08741888046264648
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,4,32,balanced,0.11583744049072267
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,4,32,balanced,0.13858688354492188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,4,32,balanced,0.19199487686157227
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,4,32,power_law_1.01,0.037052159309387205
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,4,32,balanced,0.239052791595459
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,4,32,balanced,0.03583103895187378
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,4,32,balanced,0.03857919931411743
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,2,64,power_law_1.2,0.8546189117431642
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,4,32,power_law_1.01,0.037064960002899175
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,4,32,power_law_1.01,0.03770368099212647
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,4,32,power_law_1.01,0.05601408004760742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,4,32,power_law_1.01,0.05394559860229492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,4,32,power_law_1.01,0.059216642379760744
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,4,32,power_law_1.01,0.06088064193725586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,4,32,balanced,0.14138367652893066
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,4,32,power_law_1.01,0.06260479927062988
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,4,32,power_law_1.01,0.06760191917419434
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,4,32,power_law_1.01,0.06820223808288575
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,4,32,power_law_1.01,0.07136640071868897
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,4,32,power_law_1.01,0.06893184185028076
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,4,32,power_law_1.01,0.07172607898712158
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,4,32,power_law_1.01,0.0807487964630127
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,4,32,power_law_1.01,0.03697920083999634
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,4,32,power_law_1.01,0.08909695625305177
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,4,32,power_law_1.01,0.0982694435119629
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,4,32,power_law_1.01,0.11628288269042968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,4,32,power_law_1.01,0.16158079147338866
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,4,32,power_law_1.01,0.1990336036682129
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,4,32,power_law_1.01,0.05952383995056152
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,4,32,power_law_1.01,0.26977664947509766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,4,32,power_law_1.01,0.3604902267456055
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,4,32,power_law_1.01,0.023907840251922607
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,4,32,power_law_1.01,0.0686361598968506
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,4,32,power_law_1.01,0.024285440444946287
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,4,32,power_law_1.01,0.5516735839843749
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,4,32,power_law_1.01,0.024268798828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,4,32,power_law_1.01,0.02462847948074341
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,4,32,power_law_1.01,0.024728319644927978
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,4,32,power_law_1.01,0.035651841163635255
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,4,32,power_law_1.01,0.7611302185058594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,4,32,power_law_1.01,0.03674240112304687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,4,32,power_law_1.01,0.03796864032745361
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,4,32,power_law_1.01,0.03929728031158447
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,4,32,power_law_1.01,0.045539841651916504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,4,32,power_law_1.01,0.04386176109313965
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,4,32,power_law_1.01,0.04454271793365479
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,4,32,power_law_1.01,0.05774208068847656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,4,32,power_law_1.01,0.04988671779632568
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,4,32,power_law_1.01,0.05508992195129394
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,4,32,power_law_1.01,0.06178815841674805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,4,32,power_law_1.01,0.07437439918518066
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,4,32,power_law_1.01,0.0952076816558838
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,4,32,power_law_1.01,0.12562047958374023
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,4,32,power_law_1.01,0.14599295616149904
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,4,32,power_law_1.01,0.2005081558227539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,4,32,power_law_1.01,0.03795711994171143
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,4,32,power_law_1.01,0.28380287170410157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,4,32,power_law_1.2,0.03651839971542358
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,4,32,power_law_1.01,0.04124159812927246
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,4,32,power_law_1.01,0.45139328002929685
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,4,32,power_law_1.01,0.050913281440734856
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,4,32,power_law_1.2,0.03671295881271362
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,4,32,power_law_1.2,0.03718656063079834
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,4,32,power_law_1.2,0.037285120487213136
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,4,32,power_law_1.01,0.5338496017456055
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,4,32,power_law_1.2,0.053160958290100105
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,4,32,power_law_1.2,0.0580620813369751
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,4,32,power_law_1.2,0.06153215885162353
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,4,32,power_law_1.2,0.06204160213470459
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,4,32,power_law_1.2,0.06686848163604736
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,4,32,power_law_1.2,0.06908671855926514
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,4,32,power_law_1.2,0.07192704200744629
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,4,32,power_law_1.2,0.07084799766540527
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,4,32,power_law_1.2,0.06763519763946532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,4,32,power_law_1.2,0.06973055839538575
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,4,32,power_law_1.2,0.07619711875915527
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,4,32,power_law_1.2,0.0851046371459961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,4,32,power_law_1.2,0.09449728012084961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,4,32,power_law_1.2,0.11694080352783202
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,4,32,power_law_1.2,0.136875524520874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,4,32,power_law_1.2,0.20708864212036132
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,4,32,power_law_1.2,0.2510617637634277
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,4,32,power_law_1.01,0.04581888198852539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,4,32,power_law_1.2,0.4084236907958984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,4,32,power_law_1.2,0.4559731292724609
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,4,32,power_law_1.2,0.059613437652587885
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,4,32,power_law_1.2,0.024128000736236572
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,4,32,power_law_1.2,0.024538879394531248
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,4,32,power_law_1.2,0.7002726745605469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,4,32,power_law_1.2,0.024535040855407714
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,4,32,power_law_1.2,0.03506688117980957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,4,32,power_law_1.2,0.035910398960113527
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,4,32,power_law_1.2,0.038693120479583745
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,4,32,power_law_1.2,0.03772799968719483
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,4,32,power_law_1.2,0.036943359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,4,32,power_law_1.2,1.0038848114013672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,4,32,power_law_1.2,0.0417190408706665
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,4,32,power_law_1.2,0.04581376075744629
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,4,32,power_law_1.2,0.048299517631530765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,4,32,power_law_1.2,0.045145602226257325
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,4,32,power_law_1.2,0.04482048034667969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,4,32,power_law_1.2,0.04667520046234131
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,4,32,power_law_1.2,0.05036416053771973
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,4,32,power_law_1.2,0.058800640106201174
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,4,32,power_law_1.2,0.02379136085510254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,4,32,power_law_1.2,0.05566592216491699
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,4,32,power_law_1.2,0.06679808139801026
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,4,32,power_law_1.2,0.0785433578491211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,4,32,power_law_1.2,0.1138265609741211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,4,32,power_law_1.2,0.14819071769714356
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,4,32,power_law_1.2,0.17375999450683594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,8,16,balanced,0.019688960313796997
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,4,32,power_law_1.2,0.3143244743347168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,8,16,balanced,0.019467519521713255
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,4,32,power_law_1.2,0.34856319427490234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,4,32,power_law_1.2,0.041326079368591305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,8,16,balanced,0.019900159835815428
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,8,16,balanced,0.021164801120758057
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,8,16,balanced,0.028954880237579344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,4,32,power_law_1.2,0.5340774536132813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,8,16,balanced,0.03979903936386108
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,8,16,balanced,0.061249279975891115
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,8,16,balanced,0.06410496234893799
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,8,16,balanced,0.06670335769653321
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,8,16,balanced,0.06898687839508057
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,8,16,balanced,0.07255680084228515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,4,32,power_law_1.2,0.680865249633789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,8,16,balanced,0.06487679958343506
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,8,16,balanced,0.06706687927246094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,8,16,balanced,0.06832511901855469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,8,16,balanced,0.07132927894592285
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,8,16,balanced,0.07650176048278809
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,8,16,balanced,0.09516287803649902
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,8,16,balanced,0.10647040367126466
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,8,16,balanced,0.13374208450317382
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,8,16,balanced,0.15435008049011228
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,8,16,balanced,0.20352640151977539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,8,16,balanced,0.24875263214111326
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,8,16,balanced,0.022211840152740477
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,8,16,balanced,0.02209536075592041
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,8,16,balanced,0.3656063842773437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,8,16,balanced,0.022693119049072265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,8,16,balanced,0.06074751853942871
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,8,16,balanced,0.47258239746093744
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,8,16,balanced,0.022842879295349124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,8,16,balanced,0.06333824157714843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,8,16,balanced,0.024892160892486574
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,8,16,balanced,0.029832959175109863
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,8,16,balanced,0.04414591789245605
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,8,16,balanced,0.04533887863159179
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,8,16,balanced,0.04923264026641846
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,8,16,balanced,0.04969600200653076
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,8,16,balanced,0.0506982421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,8,16,balanced,0.04452864170074463
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,8,16,balanced,0.04828288078308106
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,8,16,balanced,0.04958079814910889
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,8,16,balanced,0.051720957756042484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,8,16,balanced,0.055359997749328614
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,8,16,balanced,0.06182144165039063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,8,16,balanced,0.0735206413269043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,8,16,balanced,0.08492287635803222
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,8,16,balanced,0.11030143737792968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,8,16,balanced,0.13214719772338868
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,8,16,balanced,0.18464895248413088
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,8,16,balanced,0.2264678382873535
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,8,16,power_law_1.01,0.021413118839263917
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,8,16,balanced,0.3236761474609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,8,16,power_law_1.01,0.03731712102890015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,8,16,balanced,0.4174489593505859
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,8,16,power_law_1.01,0.03847039937973022
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,8,16,power_law_1.01,0.036866559982299804
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,8,16,power_law_1.01,0.03904639959335327
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,8,16,balanced,0.04436736106872559
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,8,16,power_law_1.01,0.05368832111358642
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,8,16,power_law_1.01,0.05743103981018066
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,8,16,power_law_1.01,0.06087039947509766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,8,16,power_law_1.01,0.06272384166717529
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,8,16,power_law_1.01,0.0666815996170044
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,8,16,power_law_1.01,0.07018239974975586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,8,16,power_law_1.01,0.07127168178558349
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,8,16,power_law_1.01,0.06764416217803955
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,8,16,power_law_1.01,0.06814976215362549
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,8,16,power_law_1.01,0.06980480194091797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,8,16,power_law_1.01,0.0735091209411621
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,8,16,power_law_1.01,0.08332927703857422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,8,16,power_law_1.01,0.09139967918395996
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,8,16,power_law_1.01,0.10692352294921874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,8,16,power_law_1.01,0.1287104034423828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,8,16,power_law_1.01,0.15534591674804688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,8,16,power_law_1.01,0.18384256362915039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,8,16,power_law_1.01,0.2848678398132324
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,8,16,balanced,0.04616064071655274
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,8,16,power_law_1.01,0.023676159381866454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,8,16,power_law_1.01,0.37601280212402344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,8,16,power_law_1.01,0.05831552028656006
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,8,16,balanced,0.08183679580688477
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,8,16,power_law_1.01,0.0268339204788208
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,8,16,power_law_1.01,0.02749696016311646
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,8,16,power_law_1.01,0.5588147354125976
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,8,16,power_law_1.01,0.028733439445495605
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,8,16,power_law_1.01,0.02972543954849243
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,8,16,power_law_1.01,0.03905280113220215
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,8,16,power_law_1.01,0.04325247764587402
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,8,16,power_law_1.01,0.04411136150360108
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,8,16,power_law_1.01,0.045250558853149415
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,8,16,power_law_1.01,0.04941696166992188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,8,16,power_law_1.01,0.048968958854675296
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,8,16,power_law_1.01,0.05543424129486084
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,8,16,power_law_1.01,0.04851456165313721
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,8,16,power_law_1.01,0.04891007900238037
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,8,16,power_law_1.01,0.050991358757019045
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,8,16,power_law_1.01,0.06256896018981933
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,8,16,power_law_1.01,0.06875008106231689
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,8,16,power_law_1.01,0.08548095703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,8,16,power_law_1.01,0.09558527946472169
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,8,16,power_law_1.01,0.13936896324157716
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,8,16,power_law_1.01,0.17938175201416015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,8,16,power_law_1.01,0.24621568679809572
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,8,16,power_law_1.01,0.6745203399658204
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,8,16,power_law_1.01,0.32607872009277344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,8,16,power_law_1.01,0.04650112152099609
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,8,16,power_law_1.2,0.02164992094039917
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,8,16,power_law_1.2,0.03732223987579346
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,8,16,power_law_1.01,0.4808204650878906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,8,16,power_law_1.2,0.03638783931732178
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,8,16,power_law_1.2,0.038353919982910156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,8,16,power_law_1.2,0.03741695880889893
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,8,16,power_law_1.01,0.053711361885070794
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,8,16,power_law_1.01,0.6113932800292969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,8,16,power_law_1.2,0.054009599685668944
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,8,16,power_law_1.2,0.06010496139526367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,8,16,power_law_1.2,0.05812863826751709
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,8,16,power_law_1.2,0.06200575828552246
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,8,16,power_law_1.2,0.06789504051208496
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,8,16,power_law_1.2,0.07002111911773681
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,8,16,power_law_1.2,0.06903552055358887
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,8,16,power_law_1.2,0.06598271846771239
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,8,16,power_law_1.2,0.07150464057922364
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,8,16,power_law_1.2,0.07700223922729492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,8,16,power_law_1.2,0.08601984024047851
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,8,16,power_law_1.2,0.10080767631530763
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,8,16,power_law_1.2,0.11525888442993164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,8,16,power_law_1.2,0.1349350357055664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,8,16,power_law_1.2,0.17960832595825196
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,8,16,power_law_1.2,0.24013952255249021
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,8,16,power_law_1.2,0.34196224212646487
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,8,16,power_law_1.2,0.023850240707397462
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,8,16,power_law_1.2,0.43222911834716793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,8,16,power_law_1.2,0.026996479034423825
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,8,16,power_law_1.2,0.05821568012237549
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,8,16,power_law_1.2,0.027159039974212644
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,8,16,power_law_1.2,0.02765183925628662
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,8,16,power_law_1.2,0.6668774414062499
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,8,16,power_law_1.2,0.029406719207763672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,8,16,power_law_1.2,0.0682585620880127
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,8,16,power_law_1.2,0.043265280723571775
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,8,16,power_law_1.2,0.04440959930419922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,8,16,power_law_1.2,0.04493440151214599
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,8,16,power_law_1.2,0.8422988891601563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,8,16,power_law_1.2,0.05032832145690917
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,8,16,power_law_1.2,0.05544320106506347
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,8,16,power_law_1.2,0.04872960090637207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,8,16,power_law_1.2,0.04850048065185547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,8,16,power_law_1.2,0.04959871768951416
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,8,16,power_law_1.2,0.05724671840667724
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,8,16,power_law_1.2,0.06559103965759278
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,8,16,power_law_1.2,0.07361663818359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,8,16,power_law_1.2,0.08797696113586426
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,8,16,power_law_1.2,0.11229311943054199
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,8,16,power_law_1.2,0.14356351852416993
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,8,16,power_law_1.2,0.1684147262573242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,8,16,power_law_1.2,0.2937932777404785
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,16,8,balanced,0.020117759704589844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,8,16,power_law_1.2,0.04208896160125732
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,8,16,power_law_1.2,0.3803776168823242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,16,8,balanced,0.02094208002090454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,16,8,balanced,0.021904640197753907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,16,8,balanced,0.02335360050201416
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,8,16,power_law_1.2,0.04710271835327148
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,16,8,balanced,0.03325439929962158
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,16,8,balanced,0.06699647903442382
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,16,8,balanced,0.06824448108673095
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,16,8,balanced,0.07164159774780274
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,16,8,balanced,0.07392896175384521
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,16,8,balanced,0.06595583915710448
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,8,16,power_law_1.2,0.0525491189956665
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,16,8,balanced,0.06812160015106201
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,16,8,balanced,0.07300992012023926
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,16,8,balanced,0.07467648029327392
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,16,8,balanced,0.07712512016296387
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,16,8,balanced,0.08200575828552245
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,16,8,balanced,0.09459456443786621
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,16,8,balanced,0.10517120361328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,16,8,balanced,0.1274662399291992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,16,8,balanced,0.1491635227203369
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,16,8,balanced,0.19649408340454103
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,16,8,balanced,0.2380287933349609
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,16,8,balanced,0.32800128936767575
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,8,16,power_law_1.2,0.5291481781005859
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,16,8,balanced,0.4174822235107422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,16,8,balanced,0.022807040214538575
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,8,16,power_law_1.2,0.7174233245849609
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,16,8,balanced,0.6232307052612305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,16,8,balanced,0.02349695920944214
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,16,8,balanced,0.025537281036376952
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,16,8,balanced,0.028768000602722166
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,16,8,balanced,0.035354878902435306
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,16,8,balanced,0.05062911987304688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,16,8,balanced,0.05414400100708008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,16,8,balanced,0.820043487548828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,16,8,balanced,0.05454720020294189
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,16,8,balanced,0.059333119392395016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,16,8,balanced,0.052255997657775874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,16,8,balanced,0.053661441802978514
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,16,8,balanced,0.05479423999786377
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,16,8,balanced,0.05742464065551758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,16,8,balanced,0.05840256214141846
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,16,8,balanced,0.061054720878601074
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,16,8,balanced,0.06536831855773925
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,16,8,balanced,0.04714240074157715
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,16,8,balanced,0.08949631690979004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,16,8,balanced,0.11229311943054199
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,16,8,balanced,0.13484800338745118
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,16,8,balanced,0.022958080768585205
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,16,8,balanced,0.06969215869903564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,16,8,balanced,0.21794944763183594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,16,8,balanced,0.31814655303955075
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,16,8,power_law_1.01,0.021713919639587402
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,16,8,balanced,0.40036094665527344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,16,8,power_law_1.01,0.028614399433135984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,16,8,balanced,0.5837798309326172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,16,8,power_law_1.01,0.0343667197227478
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,16,8,power_law_1.01,0.03764607906341553
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,16,8,balanced,0.7686988830566406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,16,8,power_law_1.01,0.041621761322021486
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,16,8,balanced,0.0783193588256836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,16,8,power_law_1.01,0.06082943916320801
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,16,8,power_law_1.01,0.06407040119171142
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,16,8,power_law_1.01,0.06607999801635742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,16,8,power_law_1.01,0.0719974422454834
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,16,8,power_law_1.01,0.06809599876403809
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,16,8,power_law_1.01,0.0690944004058838
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,16,8,power_law_1.01,0.07147136211395264
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,16,8,power_law_1.01,0.07318528175354004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,16,8,power_law_1.01,0.07726336002349854
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,16,8,power_law_1.01,0.08830464363098145
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,16,8,power_law_1.01,0.09635711669921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,16,8,power_law_1.01,0.11521663665771484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,16,8,power_law_1.01,0.1348300838470459
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,16,8,power_law_1.01,0.18076160430908203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,16,8,power_law_1.01,0.21607168197631835
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,16,8,power_law_1.01,0.0420684814453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,16,8,power_law_1.01,0.27875328063964844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,16,8,power_law_1.01,0.06965248107910156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,16,8,balanced,0.17849599838256835
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,16,8,power_law_1.01,0.02408576011657715
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,16,8,power_law_1.01,0.3955059051513672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,16,8,power_law_1.01,0.026944000720977784
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,16,8,power_law_1.01,0.02849663972854614
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,16,8,power_law_1.01,0.03077375888824463
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,16,8,power_law_1.01,0.03316351890563965
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,16,8,power_law_1.01,0.6346688079833984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,16,8,power_law_1.01,0.04958079814910889
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,16,8,power_law_1.01,0.03644799947738647
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,16,8,power_law_1.01,0.04963840007781982
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,16,8,power_law_1.01,0.7743961334228515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,16,8,power_law_1.01,0.050584321022033696
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,16,8,power_law_1.01,0.05230591773986817
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,16,8,power_law_1.01,0.05864831924438476
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,16,8,power_law_1.01,0.05217535972595215
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,16,8,power_law_1.01,0.053711361885070794
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,16,8,power_law_1.01,0.05647359848022461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,16,8,power_law_1.01,0.05991040229797363
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,16,8,power_law_1.01,0.07153151988983154
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,16,8,power_law_1.01,0.08339200019836426
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,16,8,power_law_1.01,0.1031270408630371
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,16,8,power_law_1.01,0.12329983711242676
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,16,8,power_law_1.01,0.06814208030700683
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,16,8,power_law_1.01,0.20253824234008788
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,16,8,power_law_1.01,0.2729408073425293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,16,8,power_law_1.2,0.022099199295043944
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,16,8,power_law_1.01,0.3644339370727539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,16,8,power_law_1.2,0.027527680397033693
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,16,8,power_law_1.2,0.03300352096557617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,16,8,power_law_1.01,0.058167037963867185
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,16,8,power_law_1.2,0.03586431980133057
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,16,8,power_law_1.01,0.05838719844818115
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,16,8,power_law_1.01,0.7232281494140624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,16,8,power_law_1.2,0.04168960094451905
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,16,8,power_law_1.2,0.06225920200347901
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,16,8,power_law_1.2,0.06361472129821777
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,16,8,power_law_1.2,0.06531455993652344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,16,8,power_law_1.2,0.06833792209625245
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,16,8,power_law_1.2,0.07142911911010742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,16,8,power_law_1.2,0.06684160232543945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,16,8,power_law_1.2,0.06560768127441406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,16,8,power_law_1.2,0.06838016033172607
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,16,8,power_law_1.2,0.07087999820709229
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,16,8,power_law_1.2,0.0737446403503418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,16,8,power_law_1.2,0.07874688148498535
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,16,8,power_law_1.2,0.08951040267944335
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,16,8,power_law_1.2,0.09903488159179688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,16,8,power_law_1.2,0.12585087776184084
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,16,8,power_law_1.01,0.49657215118408204
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,16,8,power_law_1.2,0.19286144256591797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,16,8,power_law_1.2,0.23848192214965822
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,16,8,power_law_1.2,0.0421017599105835
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,16,8,power_law_1.2,0.33837440490722653
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,16,8,power_law_1.01,0.1605273628234863
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,16,8,power_law_1.2,0.024207360744476318
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,16,8,power_law_1.2,0.02721152067184448
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,16,8,power_law_1.2,0.6548889923095703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,16,8,power_law_1.2,0.027678720951080323
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,16,8,power_law_1.2,0.029424641132354733
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,16,8,power_law_1.2,0.03390464067459106
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,16,8,power_law_1.2,0.04817791938781738
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,16,8,power_law_1.2,0.14463871955871582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,16,8,power_law_1.2,0.03325952053070068
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,16,8,power_law_1.2,0.8789900970458984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,16,8,power_law_1.2,0.052922878265380856
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,16,8,power_law_1.2,0.05215104103088379
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,16,8,power_law_1.2,0.05568511962890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,16,8,power_law_1.2,0.05198463916778564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,16,8,power_law_1.2,0.05435904026031494
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,16,8,power_law_1.2,0.05523839950561523
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,16,8,power_law_1.2,0.056929278373718264
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,16,8,power_law_1.2,0.40929409027099606
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,16,8,power_law_1.2,0.06299520015716553
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,16,8,power_law_1.2,0.07483007907867431
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,16,8,power_law_1.2,0.08268159866333008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,16,8,power_law_1.2,0.10381312370300293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,16,8,power_law_1.2,0.13151488304138184
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,16,8,power_law_1.2,0.1780352020263672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,16,8,power_law_1.2,0.21532415390014648
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,16,8,power_law_1.2,0.05051136016845703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,32,4,balanced,0.019916800260543824
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,32,4,balanced,0.020184319019317627
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,16,8,power_law_1.2,0.38237953186035156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,32,4,balanced,0.021943039894104004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,16,8,power_law_1.2,0.058200321197509765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,32,4,balanced,0.03379071950912475
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,32,4,balanced,0.04763775825500488
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,32,4,balanced,0.049975042343139646
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,32,4,balanced,0.06566783905029297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,32,4,balanced,0.06654848098754883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,32,4,balanced,0.06872447967529297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,16,8,power_law_1.2,0.7970111846923829
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,32,4,balanced,0.07089151859283446
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,32,4,balanced,0.07228928089141846
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,32,4,balanced,0.07550591945648193
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,32,4,balanced,0.07969664096832275
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,32,4,balanced,0.07711999893188477
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,32,4,balanced,0.08337663650512696
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,32,4,balanced,0.09420031547546387
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,32,4,balanced,0.10500479698181153
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,32,4,balanced,0.12714879989624023
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,16,8,power_law_1.2,0.30448896408081055
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,32,4,balanced,0.19636735916137696
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,16,8,power_law_1.2,0.6371161651611328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,32,4,balanced,0.2381952095031738
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,32,4,balanced,0.024888319969177248
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,32,4,balanced,0.3292825698852539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,32,4,balanced,0.049235200881958006
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,32,4,balanced,0.022359039783477783
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,32,4,balanced,0.02263040065765381
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,32,4,balanced,0.42984191894531254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,32,4,balanced,0.02399104118347168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,32,4,balanced,0.03221503973007202
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,32,4,balanced,0.027302401065826414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,32,4,balanced,0.044119038581848145
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,32,4,balanced,0.04420991897583008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,32,4,balanced,0.04536320209503174
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,32,4,balanced,0.6440383911132812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,32,4,balanced,0.06325503826141357
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,32,4,balanced,0.06378111839294434
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,32,4,balanced,0.06463615894317627
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,32,4,balanced,0.06946688175201417
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,32,4,balanced,0.07150720119476318
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,32,4,balanced,0.07310207843780517
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,32,4,balanced,0.0747711992263794
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,32,4,balanced,0.07936768054962158
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,32,4,balanced,0.09177472114562987
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,32,4,balanced,0.14782976150512694
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,32,4,balanced,0.12478336334228515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,32,4,balanced,0.1425868797302246
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,32,4,balanced,0.18672000885009765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,32,4,balanced,0.2253055953979492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,32,4,balanced,0.3342655944824219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,32,4,power_law_1.01,0.020666880607604982
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,32,4,balanced,0.4156262588500977
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,32,4,power_law_1.01,0.02286207914352417
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,32,4,power_law_1.01,0.03198463916778564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,32,4,balanced,0.6208563232421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,32,4,power_law_1.01,0.03643903970718384
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,32,4,balanced,0.06731776237487794
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,32,4,power_law_1.01,0.03831808090209961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,32,4,balanced,0.8125478363037109
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,32,4,power_law_1.01,0.04553855895996094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,32,4,power_law_1.01,0.04642047882080078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,32,4,power_law_1.01,0.06232831954956055
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,32,4,balanced,0.10171263694763184
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,32,4,power_law_1.01,0.061873922348022456
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,32,4,power_law_1.01,0.06545536041259765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,32,4,power_law_1.01,0.06910848140716552
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,32,4,power_law_1.01,0.06798975944519044
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,32,4,power_law_1.01,0.072325119972229
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,32,4,power_law_1.01,0.0737779188156128
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,32,4,power_law_1.01,0.07661952018737793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,32,4,power_law_1.01,0.08296704292297363
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,32,4,power_law_1.01,0.09366527557373047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,32,4,power_law_1.01,0.10619520187377929
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,32,4,power_law_1.01,0.12874367713928223
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,32,4,power_law_1.01,0.15105024337768555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,32,4,power_law_1.01,0.1906278419494629
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,32,4,power_law_1.01,0.2426355171203613
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,32,4,power_law_1.01,0.3428275299072266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,32,4,power_law_1.01,0.04052864074707031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,32,4,balanced,0.8416690826416016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,32,4,power_law_1.01,0.023388159275054932
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,32,4,power_law_1.01,0.4624448013305664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,32,4,power_law_1.01,0.026035199165344237
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,32,4,power_law_1.01,0.029282560348510744
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,32,4,power_law_1.01,0.03186559915542603
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,32,4,power_law_1.01,0.6622502136230468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,32,4,power_law_1.01,0.03773695945739746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,32,4,power_law_1.01,0.04290048122406006
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,32,4,power_law_1.01,0.043761920928955075
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,32,4,power_law_1.01,0.05833343982696533
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,32,4,power_law_1.01,0.8705958557128906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,32,4,power_law_1.01,0.059180798530578616
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,32,4,power_law_1.01,0.06077311992645264
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,32,4,power_law_1.01,0.06235392093658447
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,32,4,power_law_1.01,0.06441343784332275
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,32,4,power_law_1.01,0.06743167877197266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,32,4,power_law_1.01,0.0683135986328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,32,4,power_law_1.01,0.07074944019317628
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,32,4,power_law_1.01,0.08783103942871093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,32,4,power_law_1.01,0.10202367782592774
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,32,4,power_law_1.01,0.12732288360595703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,32,4,power_law_1.01,0.15262720108032227
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,32,4,power_law_1.01,0.1987980842590332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,32,4,power_law_1.01,0.2563520050048828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,32,4,power_law_1.01,0.03606784105300903
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,32,4,power_law_1.2,0.02065279960632324
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,32,4,power_law_1.01,0.359552001953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,32,4,power_law_1.2,0.022798080444335938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,32,4,power_law_1.01,0.44333438873291015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,32,4,power_law_1.2,0.03515647888183594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,32,4,power_law_1.2,0.03816319942474365
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,32,4,power_law_1.2,0.041301760673522946
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,32,4,power_law_1.01,0.6603494262695312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,32,4,power_law_1.2,0.04596992015838623
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,32,4,power_law_1.2,0.046060800552368164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,32,4,power_law_1.01,0.07568895816802979
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,32,4,power_law_1.2,0.061495041847229
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,32,4,power_law_1.01,0.8409983825683593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,32,4,power_law_1.2,0.06409471988677978
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,32,4,power_law_1.2,0.06644864082336426
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,32,4,power_law_1.2,0.06935296058654786
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,32,4,power_law_1.2,0.07250944137573243
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,32,4,power_law_1.2,0.07780608177185058
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,32,4,power_law_1.2,0.09593471527099609
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,32,4,power_law_1.2,0.10722687721252441
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,32,4,power_law_1.2,0.13169535636901855
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,32,4,power_law_1.2,0.15372544288635254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,32,4,power_law_1.2,0.200313606262207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,32,4,power_law_1.2,0.24792064666748045
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,32,4,power_law_1.2,0.030648319721221923
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,32,4,power_law_1.2,0.34749183654785154
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,32,4,power_law_1.2,0.023297278881072997
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,32,4,power_law_1.2,0.4729510498046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,32,4,power_law_1.2,0.06231808185577392
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,32,4,power_law_1.2,0.026026239395141603
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,32,4,power_law_1.2,0.07447296142578125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,32,4,power_law_1.2,0.08422400474548339
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,32,4,power_law_1.2,0.030064640045166013
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,32,4,power_law_1.2,0.03397119998931884
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,32,4,power_law_1.2,0.0422105598449707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,32,4,power_law_1.2,0.04224127769470215
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,32,4,power_law_1.2,0.059402241706848144
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,32,4,power_law_1.2,0.05992320060729981
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,32,4,power_law_1.2,0.06215551853179931
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,32,4,power_law_1.2,0.06471936225891113
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,32,4,power_law_1.2,0.06822400093078614
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,32,4,power_law_1.2,0.07015168190002441
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,32,4,power_law_1.2,0.07161856174468995
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,32,4,power_law_1.2,0.7244953918457031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,32,4,power_law_1.2,0.07753471851348877
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,32,4,power_law_1.2,0.027601919174194335
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,32,4,power_law_1.2,0.10259455680847167
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,32,4,power_law_1.2,0.1317734432220459
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,32,4,power_law_1.2,0.15546496391296388
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,32,4,power_law_1.2,0.03873280048370361
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,32,4,power_law_1.2,0.2064076805114746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,32,4,power_law_1.2,0.057368321418762205
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,32,4,power_law_1.2,0.2570124816894531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,1,128,balanced,0.02303999900817871
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,1,128,balanced,0.023514881134033203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,32,4,power_law_1.2,0.3658982467651367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,1,128,balanced,0.02333184003829956
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,32,4,power_law_1.2,0.4750105667114258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,1,128,balanced,0.023065600395202637
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,1,128,balanced,0.02351232051849365
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,1,128,balanced,0.03699968099594116
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,32,4,power_law_1.2,0.678993911743164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,32,4,power_law_1.2,0.09030783653259278
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,1,128,balanced,0.037032959461212156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,1,128,balanced,0.037095680236816406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,1,128,balanced,0.03732991933822632
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,1,128,balanced,0.0377728009223938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,1,128,balanced,0.037569279670715335
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,1,128,balanced,0.03803136110305786
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,1,128,balanced,0.0399833607673645
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,1,128,balanced,0.0413324785232544
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,32,4,power_law_1.2,0.9163839721679687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,1,128,balanced,0.040838398933410645
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,1,128,balanced,0.05146111965179443
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,1,128,balanced,0.049373440742492676
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,1,128,balanced,0.060197119712829594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,1,128,balanced,0.0530457592010498
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,1,128,balanced,0.07783423900604249
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,1,128,balanced,0.07590528011322022
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,1,128,balanced,0.08274175643920897
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,1,128,balanced,0.09538047790527344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,1,128,balanced,0.11695743560791017
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,1,128,balanced,0.1366694355010986
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,1,128,power_law_1.01,0.022777600288391115
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,1,128,power_law_1.01,0.023435521125793456
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,1,128,power_law_1.01,0.0235532808303833
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,1,128,power_law_1.01,0.03700223922729492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,1,128,power_law_1.01,0.037242879867553716
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,32,4,power_law_1.2,0.887283172607422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,1,128,power_law_1.01,0.03790208101272583
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,1,128,power_law_1.01,0.04024703979492188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,1,128,balanced,0.04026495933532715
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,1,128,power_law_1.01,0.0433625602722168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,1,128,power_law_1.01,0.05263360023498535
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,1,128,power_law_1.01,0.052231678962707515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,1,128,power_law_1.01,0.05301119804382324
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,1,128,balanced,0.03713279962539673
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,1,128,power_law_1.01,0.08436736106872558
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,1,128,power_law_1.01,0.10607359886169435
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,1,128,power_law_1.01,0.1333414363861084
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,1,128,power_law_1.01,0.18979328155517577
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,1,128,power_law_1.01,0.0372979211807251
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,1,128,power_law_1.01,0.21996416091918944
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,1,128,power_law_1.01,0.04209023952484131
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,1,128,power_law_1.01,0.04108287811279297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,1,128,power_law_1.01,0.3641484832763672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,1,128,power_law_1.01,0.06305408000946044
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,1,128,power_law_1.01,0.06808703899383545
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,1,128,power_law_1.01,0.45218559265136715
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,1,128,power_law_1.2,0.022709760665893555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,1,128,power_law_1.01,0.07661312103271485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,1,128,power_law_1.2,0.023519999980926513
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,1,128,power_law_1.01,0.680716781616211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,1,128,power_law_1.2,0.037245440483093264
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,1,128,power_law_1.2,0.03746687889099121
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,1,128,power_law_1.2,0.03732480049133301
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,1,128,power_law_1.2,0.03792383909225464
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,1,128,power_law_1.2,0.048697600364685054
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,1,128,power_law_1.2,0.040572161674499514
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,1,128,power_law_1.2,0.04690688133239746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,1,128,power_law_1.2,0.05073279857635497
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,1,128,power_law_1.2,0.05240575790405273
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,1,128,power_law_1.01,0.07667327880859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,1,128,power_law_1.2,0.05353087902069091
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,1,128,power_law_1.2,0.06815743923187256
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,1,128,power_law_1.2,0.07133696079254151
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,1,128,power_law_1.01,0.9712960052490235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,1,128,power_law_1.2,0.07642111778259278
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,1,128,power_law_1.2,0.07547135829925537
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,1,128,power_law_1.2,0.0839628791809082
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,1,128,power_law_1.2,0.1078643226623535
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,1,128,power_law_1.2,0.15254528045654298
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,1,128,power_law_1.2,0.19567487716674806
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,1,128,power_law_1.2,0.2834726333618164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,2,64,balanced,0.019942400455474855
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,1,128,power_law_1.2,0.4183705520629883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,1,128,power_law_1.2,0.02350208044052124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,2,64,balanced,0.019701759815216064
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,2,64,balanced,0.02068351984024048
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,1,128,power_law_1.2,0.6288883209228515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,2,64,balanced,0.029862399101257327
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,2,64,balanced,0.02995712041854858
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,2,64,balanced,0.03028480052947998
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,2,64,balanced,0.03038464069366455
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,2,64,balanced,0.03048703908920288
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,2,64,balanced,0.030464000701904297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,2,64,balanced,0.0340224003791809
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,2,64,balanced,0.031000320911407468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,2,64,balanced,0.03186559915542603
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,2,64,balanced,0.039957759380340574
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,2,64,balanced,0.03542144060134887
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,2,64,balanced,0.04534016132354736
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,2,64,balanced,0.036856319904327396
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,2,64,balanced,0.03984256029129028
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,2,64,balanced,0.04600704193115234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,2,64,balanced,0.06388607978820801
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,2,64,balanced,0.060756478309631344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,2,64,balanced,0.0638323211669922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,2,64,balanced,0.06960896015167237
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,2,64,balanced,0.08249600410461426
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,2,64,balanced,0.020138239860534667
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,2,64,balanced,0.12113920211791993
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,2,64,balanced,0.1528115177154541
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,2,64,power_law_1.01,0.020838398933410645
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,2,64,power_law_1.01,0.02090111970901489
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,2,64,power_law_1.01,0.020771839618682862
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,2,64,power_law_1.01,0.03004544019699097
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,2,64,power_law_1.01,0.030341119766235353
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,2,64,power_law_1.01,0.030615038871765136
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,2,64,power_law_1.01,0.031636478900909426
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,2,64,power_law_1.01,0.03415679931640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,2,64,power_law_1.01,0.03630592107772827
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,2,64,power_law_1.01,0.04837887763977051
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,2,64,power_law_1.01,0.04080639839172363
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,2,64,power_law_1.01,0.04430975914001465
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,2,64,balanced,0.08979455947875976
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,2,64,power_law_1.01,0.04401152133941651
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,2,64,power_law_1.01,0.06142848014831544
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,1,128,power_law_1.2,1.3087461853027345
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,2,64,power_law_1.01,0.061480960845947265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,2,64,power_law_1.01,0.06841983795166015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,1,128,power_law_1.2,0.8133503723144532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,2,64,power_law_1.01,0.0713753604888916
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,2,64,power_law_1.01,0.08481151580810546
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,1,128,power_law_1.2,1.9909056091308592
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,2,64,power_law_1.01,0.10861184120178222
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,2,64,power_law_1.01,0.13711487770080566
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,2,64,power_law_1.01,0.18937856674194337
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,2,64,power_law_1.01,0.31388671874999996
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,2,64,power_law_1.2,0.020679678916931152
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,2,64,power_law_1.01,0.3561548614501953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,2,64,power_law_1.01,0.4997273635864258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,2,64,power_law_1.2,0.0210150408744812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,2,64,power_law_1.2,0.02106879949569702
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,2,64,power_law_1.2,0.02995327949523926
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,2,64,power_law_1.01,0.06356863975524903
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,2,64,power_law_1.2,0.030502400398254394
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,2,64,power_law_1.01,0.7430220794677734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,2,64,power_law_1.2,0.03672447919845581
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,2,64,power_law_1.2,0.03628671884536743
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,2,64,power_law_1.2,0.04493567943572998
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,2,64,power_law_1.2,0.04509312152862549
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,2,64,power_law_1.2,0.048971519470214844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,2,64,power_law_1.2,0.05521664142608642
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,2,64,power_law_1.2,0.061327362060546876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,2,64,power_law_1.2,0.06436223983764648
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,2,64,power_law_1.2,0.06387328147888184
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,2,64,power_law_1.2,0.06275328159332275
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,2,64,power_law_1.2,0.08050432205200195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,2,64,power_law_1.2,0.11032447814941407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,2,64,power_law_1.2,0.13131135940551758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,2,64,power_law_1.2,0.1903104019165039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,2,64,power_law_1.01,0.05695615768432617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,2,64,power_law_1.2,0.2840768051147461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,2,64,power_law_1.2,0.39134590148925785
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,4,32,balanced,0.018576639890670776
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,2,64,power_law_1.2,0.04636672019958496
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,4,32,balanced,0.018543360233306886
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,4,32,balanced,0.019916800260543824
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,4,32,balanced,0.027135999202728273
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,4,32,balanced,0.02745599985122681
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,4,32,balanced,0.0319270396232605
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,4,32,balanced,0.03529855966567993
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,4,32,balanced,0.033121280670166016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,4,32,balanced,0.03024768114089966
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,4,32,balanced,0.033532159328460695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,4,32,balanced,0.032074239253997806
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,4,32,balanced,0.029913599491119387
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,4,32,balanced,0.03142656087875366
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,2,64,power_law_1.2,0.032325119972229005
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,4,32,balanced,0.041359357833862305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,4,32,balanced,0.03725696086883545
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,4,32,balanced,0.01859455943107605
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,4,32,balanced,0.05821055889129638
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,4,32,balanced,0.0531763219833374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,4,32,balanced,0.058234882354736325
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,4,32,balanced,0.06417280197143554
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,4,32,balanced,0.07362815856933594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,4,32,balanced,0.08182016372680664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,4,32,balanced,0.09893888473510742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,4,32,balanced,0.1171827220916748
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,4,32,balanced,0.16211328506469727
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,4,32,balanced,0.20791551589965823
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,4,32,power_law_1.01,0.022973439693450927
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,4,32,power_law_1.01,0.020241920948028562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,2,64,power_law_1.2,1.204542694091797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,2,64,power_law_1.2,0.6402124786376954
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,2,64,power_law_1.2,0.8419123077392578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,4,32,power_law_1.01,0.027239680290222168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,4,32,power_law_1.01,0.027578880786895753
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,4,32,balanced,0.04343552112579346
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,4,32,power_law_1.01,0.033365759849548336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,4,32,power_law_1.01,0.03716480016708374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,4,32,power_law_1.01,0.034213120937347415
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,4,32,power_law_1.01,0.03579648017883301
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,4,32,power_law_1.01,0.03884543895721436
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,4,32,power_law_1.01,0.04950272083282471
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,4,32,power_law_1.01,0.05671296119689941
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,2,64,power_law_1.2,0.06987520217895507
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,4,32,power_law_1.01,0.05979135990142822
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,4,32,power_law_1.01,0.05782400131225586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,4,32,power_law_1.01,0.058740482330322266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,4,32,power_law_1.01,0.06353792190551757
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,4,32,power_law_1.01,0.07224063873291016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,4,32,power_law_1.01,0.08960127830505371
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,4,32,power_law_1.01,0.10129152297973634
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,4,32,power_law_1.01,0.1355033588409424
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,4,32,power_law_1.01,0.02159615993499756
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,4,32,power_law_1.01,0.19399040222167968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,4,32,power_law_1.01,0.26896640777587894
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,4,32,power_law_1.01,0.03330303907394409
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,4,32,power_law_1.2,0.020401918888092042
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,4,32,power_law_1.2,0.020846080780029298
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,4,32,power_law_1.01,0.04141183853149414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,4,32,power_law_1.01,0.4407014465332031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,4,32,power_law_1.2,0.02381824016571045
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,4,32,power_law_1.2,0.027235839366912838
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,4,32,power_law_1.2,0.029770240783691403
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,4,32,power_law_1.2,0.03301503896713257
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,4,32,power_law_1.01,0.6360166549682618
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,4,32,power_law_1.2,0.03179775953292847
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,4,32,power_law_1.2,0.032131841182708745
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,4,32,power_law_1.2,0.043564801216125486
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,4,32,power_law_1.2,0.0504204797744751
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,4,32,power_law_1.2,0.056897277832031246
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,4,32,power_law_1.2,0.05812736034393311
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,4,32,power_law_1.2,0.05632895946502685
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,4,32,power_law_1.2,0.05715968132019043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,4,32,power_law_1.2,0.0608460807800293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,4,32,power_law_1.2,0.06945280075073243
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,4,32,power_law_1.2,0.08089471817016601
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,4,32,power_law_1.2,0.09488639831542969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,4,32,power_law_1.2,0.12135552406311034
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,4,32,power_law_1.2,0.17079423904418944
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,4,32,power_law_1.2,0.2212416076660156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,4,32,power_law_1.2,0.03843967914581299
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,8,16,balanced,0.0199782395362854
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,4,32,power_law_1.2,0.036453120708465576
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,8,16,balanced,0.01987712025642395
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,8,16,balanced,0.021192960739135742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,8,16,balanced,0.022590720653533937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,8,16,balanced,0.05080448150634766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,4,32,power_law_1.2,0.7116531372070313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,8,16,balanced,0.04600063800811768
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,8,16,balanced,0.053118720054626464
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,8,16,balanced,0.05090559959411621
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,8,16,balanced,0.048381438255310064
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,8,16,balanced,0.054529280662536615
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,8,16,balanced,0.051928319931030274
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,8,16,balanced,0.05952127933502197
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,8,16,balanced,0.06108799934387207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,4,32,power_law_1.2,0.32783615112304687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,8,16,balanced,0.06787583827972413
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,8,16,balanced,0.05776639938354492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,8,16,balanced,0.06121088027954101
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,8,16,balanced,0.06597375869750977
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,8,16,balanced,0.08264960289001465
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,8,16,balanced,0.08981504440307617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,8,16,balanced,0.02427135944366455
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,8,16,balanced,0.11032959938049317
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,8,16,balanced,0.12338175773620605
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,8,16,balanced,0.16984832763671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,8,16,balanced,0.20528640747070312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,4,32,power_law_1.01,0.2972889518737793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,8,16,balanced,0.2836096000671387
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,8,16,power_law_1.01,0.02310784101486206
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,4,32,power_law_1.2,0.9898508453369141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,8,16,balanced,0.05466752052307129
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,8,16,balanced,0.36852863311767575
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,8,16,power_law_1.01,0.03195904016494751
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,8,16,power_law_1.01,0.031521279811859135
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,4,32,power_law_1.2,0.49263744354248046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,8,16,power_law_1.01,0.044121599197387694
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,8,16,power_law_1.01,0.046649599075317384
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,8,16,power_law_1.01,0.04940927982330322
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,8,16,power_law_1.01,0.050245118141174314
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,8,16,power_law_1.01,0.05250432014465332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,8,16,power_law_1.01,0.052157440185546876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,8,16,power_law_1.01,0.0533516788482666
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,8,16,power_law_1.01,0.058511362075805665
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,8,16,power_law_1.01,0.06246016025543213
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,8,16,power_law_1.01,0.06292863845825195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,8,16,power_law_1.01,0.06876287937164308
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,8,16,power_law_1.01,0.07817855834960938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,8,16,power_law_1.01,0.08236543655395508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,8,16,power_law_1.01,0.10452863693237305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,8,16,power_law_1.01,0.11766400337219238
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,8,16,power_law_1.01,0.15572735786437988
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,8,16,power_law_1.01,0.2103104019165039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,8,16,power_law_1.01,0.2876889610290527
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,8,16,power_law_1.01,0.34715393066406247
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,8,16,power_law_1.2,0.02302592039108276
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,8,16,power_law_1.2,0.029516799449920656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,8,16,power_law_1.2,0.030956799983978274
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,8,16,power_law_1.2,0.03158143997192383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,8,16,power_law_1.2,0.04210559844970703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,8,16,power_law_1.01,0.06533504009246827
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,8,16,power_law_1.01,0.064235520362854
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,8,16,power_law_1.2,0.04442368030548095
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,8,16,power_law_1.01,0.8108351898193359
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,8,16,power_law_1.2,0.04977536201477051
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,8,16,power_law_1.2,0.05100160121917725
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,8,16,power_law_1.2,0.052597761154174805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,8,16,power_law_1.2,0.052906241416931156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,8,16,power_law_1.2,0.05651840209960938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,8,16,power_law_1.2,0.058685441017150876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,8,16,power_law_1.2,0.06480768203735351
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,8,16,power_law_1.2,0.06585599899291991
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,8,16,power_law_1.2,0.06690303802490234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,8,16,power_law_1.2,0.06909567832946777
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,8,16,power_law_1.2,0.07302271842956543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,8,16,power_law_1.2,0.07990911960601807
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,8,16,power_law_1.01,0.0332096004486084
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,8,16,power_law_1.2,0.09575167655944825
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,8,16,power_law_1.2,0.11962495803833009
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,8,16,power_law_1.2,0.13726847648620605
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,8,16,power_law_1.01,0.5543078231811523
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,8,16,power_law_1.2,0.2786854362487793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,16,8,balanced,0.020055038928985598
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,16,8,balanced,0.02075007915496826
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,16,8,balanced,0.021790719032287596
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,16,8,balanced,0.029724159240722657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,8,16,power_law_1.2,0.3585587310791015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,8,16,power_law_1.2,0.8188326263427734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,16,8,balanced,0.05240960121154785
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,16,8,balanced,0.05395711898803711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,16,8,balanced,0.052839679718017576
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,16,8,balanced,0.055453438758850095
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,16,8,balanced,0.05923711776733398
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,16,8,balanced,0.05697535991668701
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,16,8,balanced,0.06171775817871093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,8,16,power_law_1.2,0.1939673614501953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,16,8,balanced,0.069651198387146
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,16,8,balanced,0.05968768119812011
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,16,8,balanced,0.06421631813049317
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,8,16,power_law_1.2,1.003909149169922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,16,8,balanced,0.06752511978149414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,16,8,balanced,0.07420159816741943
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,16,8,balanced,0.08208127975463866
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,16,8,balanced,0.10561920166015626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,16,8,balanced,0.1169753646850586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,16,8,balanced,0.15252479553222656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,16,8,balanced,0.1768998336791992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,16,8,balanced,0.24280576705932616
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,16,8,balanced,0.2992166328430176
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,16,8,power_law_1.01,0.022101759910583496
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,16,8,balanced,0.025564160346984867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,16,8,power_law_1.01,0.026552319526672363
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,16,8,power_law_1.01,0.03108479976654053
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,8,16,power_law_1.2,0.5009024047851562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,16,8,balanced,0.5755456161499024
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,16,8,power_law_1.01,0.0492902421951294
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,16,8,power_law_1.01,0.0521446418762207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,16,8,power_law_1.01,0.055543041229248045
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,16,8,power_law_1.01,0.0543782377243042
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,16,8,power_law_1.01,0.05907072067260742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,16,8,power_law_1.01,0.06022784233093261
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,16,8,power_law_1.01,0.06718080043792725
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,16,8,power_law_1.01,0.06723968029022218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,16,8,power_law_1.01,0.06479616165161133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,16,8,power_law_1.01,0.06533120155334474
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,16,8,power_law_1.01,0.06575104236602783
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,16,8,power_law_1.01,0.07296895980834961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,16,8,power_law_1.01,0.0822822380065918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,16,8,power_law_1.01,0.09144960403442383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,16,8,power_law_1.01,0.10823936462402343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,16,8,power_law_1.01,0.1293619155883789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,16,8,power_law_1.01,0.17727231979370117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,16,8,power_law_1.01,0.03216000080108643
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,16,8,power_law_1.01,0.03499135971069336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,16,8,power_law_1.01,0.3237427139282226
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,16,8,balanced,0.062202877998352044
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,16,8,power_law_1.2,0.022259199619293214
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,16,8,power_law_1.01,0.3761971282958984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,16,8,power_law_1.2,0.024191999435424806
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,16,8,power_law_1.2,0.03276031970977783
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,16,8,power_law_1.2,0.03367552042007446
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,16,8,power_law_1.2,0.05418367862701416
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,16,8,power_law_1.01,0.7510182189941406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,16,8,power_law_1.2,0.05237504005432129
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,16,8,power_law_1.2,0.0546457576751709
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,16,8,balanced,0.4527910232543945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,16,8,power_law_1.2,0.058305277824401855
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,16,8,power_law_1.2,0.0628876781463623
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,16,8,power_law_1.01,0.19458175659179688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,16,8,power_law_1.2,0.06845823764801026
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,16,8,power_law_1.2,0.06190847873687744
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,16,8,power_law_1.2,0.06464384078979493
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,16,8,power_law_1.2,0.06441728115081788
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,16,8,power_law_1.2,0.0740121603012085
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,16,8,power_law_1.2,0.07985663890838623
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,16,8,power_law_1.2,0.09712767601013184
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,16,8,power_law_1.2,0.030593280792236326
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,16,8,power_law_1.2,0.11691007614135743
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,16,8,power_law_1.2,0.13841407775878906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,16,8,power_law_1.01,0.587267837524414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,16,8,power_law_1.2,0.19771135330200196
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,16,8,power_law_1.2,0.053049597740173336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,16,8,power_law_1.2,0.2305407905578613
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,32,4,balanced,0.02049407958984375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,16,8,power_law_1.2,0.3438374328613281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,32,4,balanced,0.021532158851623535
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,16,8,power_law_1.2,0.06014976024627685
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,32,4,balanced,0.023376638889312747
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,16,8,power_law_1.2,0.43808513641357416
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,32,4,balanced,0.028997120857238767
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,32,4,balanced,0.036354560852050785
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,32,4,balanced,0.047873277664184574
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,32,4,balanced,0.058746881484985344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,32,4,balanced,0.05875455856323242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,32,4,balanced,0.06100863933563232
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,32,4,balanced,0.06465919971466064
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,16,8,power_law_1.2,0.6968128204345703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,32,4,balanced,0.07031424045562744
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,32,4,balanced,0.06339327812194824
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,32,4,balanced,0.06795775890350342
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,32,4,balanced,0.06803584098815918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,32,4,balanced,0.07088128089904785
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,32,4,balanced,0.07531519889831542
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,32,4,balanced,0.08450688362121582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,32,4,balanced,0.09184639930725098
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,32,4,balanced,0.12192383766174317
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,32,4,balanced,0.1366528034210205
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,16,8,power_law_1.2,0.9285465240478515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,32,4,balanced,0.18193023681640624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,32,4,balanced,0.21334272384643554
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,32,4,balanced,0.29957759857177735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,32,4,power_law_1.01,0.02190079927444458
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,32,4,balanced,0.3729164886474609
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,32,4,power_law_1.01,0.02587775945663452
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,32,4,power_law_1.01,0.029922559261322024
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,32,4,balanced,0.5644095993041992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,32,4,power_law_1.01,0.03945472002029419
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,32,4,balanced,0.06958975791931152
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,32,4,power_law_1.01,0.059448318481445314
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,32,4,balanced,0.7349043273925782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,32,4,power_law_1.01,0.06197247982025147
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,32,4,power_law_1.01,0.060084481239318845
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,32,4,power_law_1.01,0.06257919788360596
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,32,4,power_law_1.01,0.06903423786163329
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,32,4,power_law_1.01,0.06699391841888427
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,32,4,power_law_1.01,0.06763135910034179
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,32,4,power_law_1.01,0.06949888229370117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,32,4,power_law_1.01,0.07646592140197754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,32,4,power_law_1.01,0.08783616065979004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,32,4,power_law_1.01,0.09758591651916504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,32,4,power_law_1.01,0.1183296012878418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,32,4,power_law_1.01,0.034757120609283446
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,32,4,power_law_1.01,0.18617599487304687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,32,4,power_law_1.01,0.22633216857910154
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,32,4,power_law_1.01,0.3182592010498047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,32,4,power_law_1.01,0.41120128631591796
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,32,4,power_law_1.2,0.022207999229431154
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,32,4,power_law_1.01,0.044999680519104
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,32,4,power_law_1.2,0.024885759353637696
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,32,4,power_law_1.01,0.5995967864990235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,32,4,power_law_1.2,0.03568128108978271
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,32,4,power_law_1.01,0.7628928375244141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,32,4,power_law_1.2,0.046878719329833986
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,32,4,power_law_1.01,0.1423961639404297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,32,4,power_law_1.2,0.056755199432373046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,32,4,power_law_1.2,0.05842432022094727
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,32,4,power_law_1.2,0.061043200492858884
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,32,4,power_law_1.2,0.061708798408508306
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,32,4,power_law_1.2,0.06591231822967529
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,32,4,power_law_1.2,0.0725222396850586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,32,4,power_law_1.2,0.07137407779693603
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,32,4,power_law_1.2,0.0688153600692749
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,32,4,power_law_1.2,0.06731391906738281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,32,4,power_law_1.01,0.07311999797821045
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,32,4,power_law_1.2,0.06861440181732179
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,32,4,power_law_1.2,0.0760537576675415
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,32,4,power_law_1.2,0.08820608139038086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,32,4,power_law_1.2,0.10059136390686034
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,32,4,power_law_1.2,0.12447487831115722
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,32,4,power_law_1.2,0.14559359550476075
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,32,4,power_law_1.2,0.19272703170776367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,32,4,power_law_1.2,0.23246976852416995
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,1,128,balanced,0.05491456031799317
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,32,4,power_law_1.2,0.4684659194946289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,1,128,balanced,0.05462143898010254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,32,4,power_law_1.2,0.04144896030426025
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,1,128,balanced,0.05529856204986572
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,32,4,power_law_1.2,0.6567436981201171
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,1,128,balanced,0.07456128120422363
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,32,4,power_law_1.01,0.06651648044586181
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,1,128,balanced,0.1298316764831543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,1,128,balanced,0.1291443157196045
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,1,128,balanced,0.13038463592529298
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,1,128,balanced,0.13698559761047363
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,32,4,power_law_1.2,0.028755199909210206
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,1,128,balanced,0.1364249610900879
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,1,128,balanced,0.138657283782959
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,1,128,balanced,0.14296704292297363
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,1,128,balanced,0.1503116798400879
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,1,128,balanced,0.15339903831481932
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,1,128,balanced,0.16722688674926758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,1,128,balanced,0.16917760848999025
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,1,128,balanced,0.18839168548583984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,1,128,balanced,0.19410560607910157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,1,128,balanced,0.21817600250244143
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,32,4,power_law_1.2,0.3290316772460938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,1,128,balanced,0.24126720428466797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,1,128,balanced,0.3249523162841797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,1,128,balanced,0.1361190414428711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,1,128,balanced,0.37277057647705075
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,1,128,balanced,0.15353216171264647
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,1,128,balanced,0.03367808103561402
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,1,128,balanced,0.5554790496826172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,1,128,balanced,0.6686553955078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,1,128,balanced,0.0306713604927063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,1,128,balanced,0.048578557968139646
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,1,128,balanced,0.0613708782196045
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,1,128,balanced,0.06272768020629883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,1,128,balanced,0.06461056232452392
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,1,128,balanced,0.06907648086547852
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,1,128,balanced,0.06741631984710693
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,1,128,balanced,0.9733414459228517
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,1,128,balanced,0.07007999897003174
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,1,128,balanced,0.07430272102355957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,1,128,balanced,0.07840767860412598
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,1,128,balanced,0.08786944389343262
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,1,128,balanced,1.2524301147460937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,1,128,balanced,0.08261119842529296
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,1,128,balanced,0.08389887809753419
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,1,128,balanced,0.09238783836364746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,1,128,balanced,0.09866239547729491
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,1,128,balanced,0.12965888023376465
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,1,128,balanced,0.13371007919311523
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,1,128,balanced,0.03230463981628418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,1,128,balanced,0.15855232238769532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,1,128,balanced,0.22167423248291013
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,1,128,balanced,0.2635840034484863
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,1,128,balanced,0.35429630279541013
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,1,128,balanced,0.43404289245605465
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,1,128,power_law_1.01,0.08489343643188477
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,1,128,power_law_1.01,0.08476544380187988
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,1,128,balanced,0.640926742553711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,1,128,power_law_1.01,0.09480832099914552
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,1,128,balanced,0.8157107543945312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,1,128,power_law_1.01,0.12140928268432619
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,1,128,power_law_1.01,0.1241971206665039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,1,128,balanced,0.11692928314208983
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,1,128,power_law_1.01,0.13000960350036622
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,1,128,power_law_1.01,0.15017215728759764
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,1,128,power_law_1.01,0.15584383964538576
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,1,128,power_law_1.01,0.14614656448364258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,1,128,power_law_1.01,0.15477888107299803
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,1,128,power_law_1.01,0.16640640258789063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,1,128,power_law_1.01,0.18747903823852538
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,1,128,power_law_1.01,0.17900672912597657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,1,128,power_law_1.01,0.2119308853149414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,32,4,power_law_1.2,0.8412351989746094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,1,128,power_law_1.01,0.3063385581970215
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,1,128,power_law_1.01,0.3658483123779297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,1,128,power_law_1.01,0.13154047966003418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,1,128,power_law_1.01,0.13972864151000977
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,1,128,power_law_1.01,0.58866943359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,1,128,power_law_1.01,0.9417919921875001
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,1,128,power_law_1.01,0.2316736030578613
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,1,128,power_law_1.01,1.1570674896240234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,1,128,power_law_1.01,0.055696640014648434
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,1,128,power_law_1.01,0.05644032001495362
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,1,128,power_law_1.01,0.4645427322387695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,1,128,power_law_1.01,0.05736320018768311
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,1,128,power_law_1.01,0.05923456192016602
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,1,128,power_law_1.01,0.06059904098510742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,1,128,power_law_1.01,0.06655488014221192
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,1,128,power_law_1.01,1.6567283630371095
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,1,128,power_law_1.01,0.06717440128326416
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,1,128,power_law_1.01,0.07383808135986328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,1,128,power_law_1.01,0.08147839546203614
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,1,128,power_law_1.01,0.08179583549499511
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,1,128,power_law_1.01,0.08570752143859864
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,1,128,power_law_1.01,0.09275520324707032
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,1,128,power_law_1.01,0.10445440292358399
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,1,128,power_law_1.01,0.1325107192993164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,1,128,power_law_1.01,0.12026623725891113
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,1,128,power_law_1.01,2.2934156799316403
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,1,128,power_law_1.01,0.14443903923034668
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,1,128,power_law_1.01,0.20242944717407227
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,1,128,power_law_1.01,0.21532543182373046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,1,128,power_law_1.01,0.3043379211425781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,1,128,power_law_1.01,0.371855354309082
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,1,128,power_law_1.01,0.5572390365600586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,1,128,power_law_1.01,0.08740863800048829
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,1,128,power_law_1.01,0.6917068481445312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,1,128,power_law_1.2,0.08504832267761231
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,1,128,power_law_1.01,3.561931457519531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,1,128,power_law_1.2,0.08477696418762207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,1,128,power_law_1.2,0.09418496131896972
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,1,128,power_law_1.01,1.1947891235351562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,1,128,power_law_1.2,0.13570048332214354
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,1,128,power_law_1.2,0.12172927856445312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,1,128,power_law_1.2,0.13293439865112305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,1,128,power_law_1.2,0.12844160079956055
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,1,128,power_law_1.01,1.6094694519042967
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,1,128,power_law_1.2,0.1339622402191162
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,1,128,power_law_1.2,0.14888064384460448
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,1,128,power_law_1.01,4.6767602539062505
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,1,128,power_law_1.2,0.14732543945312498
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,1,128,power_law_1.2,0.16471935272216798
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,1,128,power_law_1.2,0.16002431869506836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,1,128,power_law_1.2,0.1685171127319336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,1,128,power_law_1.2,0.20036224365234373
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,1,128,power_law_1.2,0.22751487731933592
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,1,128,power_law_1.2,0.24622207641601562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,1,128,power_law_1.01,2.258117065429688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,1,128,power_law_1.2,0.34890113830566405
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,1,128,power_law_1.2,0.43711616516113283
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,1,128,power_law_1.2,0.5812902450561523
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,1,128,power_law_1.2,0.7603648376464844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,1,128,power_law_1.01,3.412126770019531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,1,128,power_law_1.2,1.3341786193847658
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,1,128,power_law_1.2,1.5807116699218748
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,1,128,power_law_1.2,0.05545216083526612
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,1,128,power_law_1.2,0.05612671852111817
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,1,128,power_law_1.2,0.05972864151000976
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,1,128,power_law_1.2,0.058767361640930174
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,1,128,power_law_1.2,0.06038400173187256
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,1,128,power_law_1.2,0.06497536182403565
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,1,128,power_law_1.2,0.06685056209564208
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,1,128,power_law_1.2,2.4582949829101564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,1,128,power_law_1.2,0.07410175800323486
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,1,128,power_law_1.2,0.07922431945800781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,1,128,power_law_1.2,0.08278783798217773
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,1,128,power_law_1.2,0.0875391960144043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,1,128,power_law_1.2,0.09541248321533204
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,1,128,power_law_1.2,0.1266598415374756
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,1,128,power_law_1.2,0.20512512207031247
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,1,128,power_law_1.2,0.14026240348815916
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,1,128,power_law_1.2,0.1362598419189453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,1,128,power_law_1.2,0.15472512245178222
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,1,128,power_law_1.2,3.21378173828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,1,128,power_law_1.2,0.2129779243469238
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,1,128,power_law_1.2,0.2642444801330567
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,1,128,power_law_1.2,0.39674751281738285
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,1,128,power_law_1.2,0.5456204986572265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,1,128,power_law_1.2,0.742402572631836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,1,128,power_law_1.2,0.959109115600586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,2,64,balanced,0.025873920917510985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,2,64,balanced,0.04917247772216797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,2,64,balanced,0.053628158569335935
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,1,128,power_law_1.2,0.10175488471984864
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,1,128,power_law_1.2,1.6727526855468748
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,2,64,balanced,0.07307263851165771
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,2,64,balanced,0.11761535644531249
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,2,64,balanced,0.12601728439331056
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,2,64,balanced,0.13017215728759765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,2,64,balanced,0.12777215957641602
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,2,64,balanced,0.14161151885986328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,2,64,balanced,0.129235200881958
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,1,128,power_law_1.2,2.291760711669922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,2,64,balanced,0.1409331226348877
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,2,64,balanced,0.14924415588378906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,2,64,balanced,0.15316864013671877
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,2,64,balanced,0.16471424102783203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,2,64,balanced,0.17075328826904296
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,2,64,balanced,0.12172800064086915
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,2,64,balanced,0.1748851203918457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,2,64,balanced,0.19203199386596678
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,2,64,balanced,0.2426214408874512
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,2,64,balanced,0.27906688690185544
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,2,64,balanced,0.13413887977600097
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,2,64,balanced,0.386693115234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,1,128,power_law_1.2,3.244268798828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,2,64,balanced,0.45255550384521487
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,2,64,balanced,0.027553279399871827
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,2,64,balanced,0.6754188537597656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,2,64,balanced,0.8425357055664062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,2,64,balanced,0.036674559116363525
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,2,64,balanced,0.05075712203979492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,2,64,balanced,0.06414591789245605
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,2,64,balanced,1.2347097778320313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,2,64,balanced,0.07121791839599609
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,2,64,balanced,0.07545216083526611
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,2,64,balanced,0.0760256004333496
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,2,64,balanced,0.08710399627685547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,2,64,balanced,0.07881216049194337
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,2,64,balanced,0.03178240060806274
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,2,64,balanced,0.08203519821166992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,2,64,balanced,0.08658304214477539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,2,64,balanced,0.09371007919311523
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,1,128,power_law_1.2,4.332476196289063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,2,64,balanced,0.09790592193603516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,1,128,power_law_1.2,8.218748168945313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,2,64,balanced,0.11564031600952149
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,2,64,balanced,0.13589759826660158
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,2,64,balanced,0.128154878616333
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,2,64,balanced,0.2033817672729492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,2,64,balanced,0.17852415084838866
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,2,64,balanced,0.28894208908081054
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,2,64,balanced,0.3517081451416016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,2,64,balanced,1.5912973022460937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,2,64,balanced,0.48626430511474605
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,1,128,power_law_1.2,5.85088134765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,2,64,balanced,0.6207807922363282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,2,64,balanced,0.10284928321838378
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,2,64,power_law_1.01,0.08018048286437988
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,2,64,power_law_1.01,0.08808704376220702
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,2,64,balanced,0.06684927940368653
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,2,64,balanced,0.9234687805175781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,2,64,power_law_1.01,0.10842623710632324
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,2,64,power_law_1.01,0.11916928291320801
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,2,64,power_law_1.01,0.1160268783569336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,2,64,power_law_1.01,0.12904319763183594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,2,64,balanced,1.1881855773925782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,2,64,power_law_1.01,0.13406975746154787
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,2,64,power_law_1.01,0.13568639755249023
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,2,64,power_law_1.01,0.14492671966552734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,2,64,power_law_1.01,0.1508454418182373
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,2,64,power_law_1.01,0.16611072540283203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,2,64,power_law_1.01,0.16981887817382812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,2,64,power_law_1.01,0.18617599487304687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,2,64,power_law_1.01,0.08289024353027344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,2,64,power_law_1.01,0.20235008239746094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,2,64,power_law_1.01,0.08567423820495605
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,2,64,power_law_1.01,0.25485567092895506
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,2,64,power_law_1.01,0.2918668746948242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,2,64,power_law_1.01,0.3956800079345703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,2,64,power_law_1.01,0.12516223907470703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,2,64,power_law_1.01,0.5017907333374023
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,2,64,power_law_1.01,0.7412287902832031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,2,64,power_law_1.01,0.05392384052276611
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,2,64,power_law_1.01,0.9529241943359376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,2,64,power_law_1.01,0.05440383911132812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,2,64,power_law_1.01,0.055013117790222166
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,2,64,power_law_1.01,0.05689856052398682
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,2,64,power_law_1.01,0.0582425594329834
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,2,64,power_law_1.01,1.429381103515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,2,64,power_law_1.01,0.06782336235046386
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,2,64,power_law_1.01,0.06326528072357178
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,2,64,power_law_1.01,0.07744256019592285
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,2,64,power_law_1.01,0.07769472122192382
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,2,64,power_law_1.01,0.08114560127258301
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,2,64,power_law_1.01,0.08149120330810547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,2,64,power_law_1.01,0.09023103713989258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,2,64,power_law_1.01,0.09433216094970703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,2,64,power_law_1.01,0.10727935791015626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,2,64,power_law_1.01,0.12055168151855469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,2,64,power_law_1.01,0.12529919624328614
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,2,64,power_law_1.01,2.004339141845703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,2,64,power_law_1.01,0.1326630401611328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,2,64,power_law_1.01,0.17919103622436522
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,2,64,power_law_1.01,0.21195392608642577
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,2,64,power_law_1.01,0.24634496688842775
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,2,64,power_law_1.01,0.30116607666015627
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,2,64,power_law_1.01,0.5071449661254883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,2,64,power_law_1.01,0.6736345672607422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,2,64,power_law_1.01,2.830751953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,2,64,power_law_1.2,0.08020223617553711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,2,64,power_law_1.2,0.08005887985229491
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,2,64,power_law_1.2,0.08539520263671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,2,64,power_law_1.01,0.9602444458007813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,2,64,power_law_1.2,0.08527615547180176
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,2,64,power_law_1.2,0.10754431724548338
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,2,64,power_law_1.2,0.11445631980895996
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,2,64,power_law_1.2,0.11792384147644044
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,2,64,power_law_1.2,0.12764672279357908
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,2,64,power_law_1.2,0.1343616008758545
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,2,64,power_law_1.2,0.13155327796936037
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,2,64,power_law_1.2,0.13984767913818358
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,2,64,power_law_1.2,0.14768768310546876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,2,64,power_law_1.01,1.7999795532226563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,2,64,power_law_1.2,0.17419136047363282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,2,64,power_law_1.2,0.1895039939880371
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,2,64,power_law_1.2,0.1965388870239258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,2,64,power_law_1.2,0.20648704528808595
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,2,64,power_law_1.2,0.2699148750305176
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,2,64,power_law_1.01,1.2461516571044922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,2,64,power_law_1.2,0.30694400787353515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,2,64,power_law_1.2,0.4937164688110352
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,2,64,power_law_1.01,2.8367730712890626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,2,64,power_law_1.2,0.6057804870605469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,2,64,power_law_1.2,0.1508518409729004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,2,64,power_law_1.01,4.320425109863281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,2,64,power_law_1.2,0.9159999847412109
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,2,64,power_law_1.2,0.054236159324646
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,2,64,power_law_1.2,0.05844863891601563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,2,64,power_law_1.2,0.055366401672363286
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,2,64,power_law_1.2,0.05686528205871581
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,2,64,power_law_1.2,1.3172454833984375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,2,64,power_law_1.2,0.05973120212554932
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,2,64,power_law_1.2,0.06334080219268798
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,2,64,power_law_1.2,0.06696447849273682
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,2,64,power_law_1.2,0.07587584018707275
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,2,64,power_law_1.2,0.07974016189575195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,2,64,power_law_1.2,0.08398976325988769
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,2,64,power_law_1.2,2.0090316772460937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,2,64,power_law_1.2,0.09148799896240234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,2,64,power_law_1.2,0.0997043228149414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,2,64,power_law_1.2,0.11006719589233398
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,2,64,power_law_1.2,0.13218432426452636
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,2,64,power_law_1.2,0.13591423988342285
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,2,64,power_law_1.2,0.19112064361572265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,2,64,power_law_1.2,0.22618240356445313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,2,64,power_law_1.2,2.6740429687500002
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,2,64,power_law_1.2,0.07683839797973632
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,2,64,power_law_1.2,0.32536193847656253
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,2,64,power_law_1.2,0.3891212844848633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,2,64,power_law_1.2,0.13022463798522949
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,2,64,power_law_1.2,0.6122060775756836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,2,64,power_law_1.2,0.8914995574951172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,4,32,balanced,0.02414463996887207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,4,32,balanced,0.04294528007507324
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,2,64,power_law_1.2,1.2669554901123046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,4,32,balanced,0.051560959815979
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,4,32,balanced,0.07422848224639893
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,4,32,balanced,0.0929036808013916
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,4,32,balanced,0.12549247741699218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,4,32,balanced,0.14106623649597166
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,4,32,balanced,0.1300723171234131
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,2,64,power_law_1.2,4.280205993652343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,4,32,balanced,0.13156864166259766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,4,32,balanced,0.14216575622558594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,4,32,balanced,0.14898688316345216
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,4,32,balanced,0.14790656089782714
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,4,32,balanced,0.15971455574035645
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,4,32,balanced,0.1667148780822754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,4,32,balanced,0.1789619255065918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,4,32,balanced,0.16594303131103516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,4,32,balanced,0.17861120223999022
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,4,32,balanced,0.20735103607177735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,4,32,balanced,0.23510784149169922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,2,64,power_law_1.2,1.5772709655761719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,4,32,balanced,0.30728960037231445
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,4,32,balanced,0.36552833557128905
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,4,32,balanced,0.5127372741699219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,4,32,balanced,0.6288755035400391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,2,64,power_law_1.2,2.631092529296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,4,32,balanced,0.02598144054412842
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,2,64,power_law_1.2,3.259467468261719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,4,32,balanced,0.028298239707946777
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,4,32,balanced,0.03519615888595581
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,4,32,balanced,0.05364480018615723
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,4,32,balanced,0.06296192169189453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,4,32,balanced,0.07375999927520752
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,4,32,balanced,1.178562545776367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,4,32,balanced,0.08711423873901367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,4,32,balanced,0.0781056022644043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,4,32,balanced,0.0850534439086914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,4,32,balanced,0.08729599952697754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,4,32,balanced,0.09394559860229493
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,4,32,balanced,0.10037887573242188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,4,32,balanced,0.11313664436340332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,4,32,balanced,0.12600831985473632
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,4,32,balanced,1.7378125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,4,32,balanced,0.11135616302490234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,4,32,balanced,0.1260979175567627
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,4,32,balanced,0.1541427230834961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,4,32,balanced,0.9272051239013672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,4,32,balanced,0.234073600769043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,4,32,balanced,0.2890931129455566
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,4,32,balanced,0.07713280200958253
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,4,32,balanced,0.4222848129272461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,4,32,balanced,0.5346060943603516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,4,32,balanced,0.7789516448974609
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,4,32,power_law_1.01,0.08465920448303223
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,4,32,balanced,0.1829120063781738
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,4,32,power_law_1.01,0.08560895919799805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,4,32,power_law_1.01,0.08492032051086426
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,4,32,balanced,2.27299072265625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,4,32,power_law_1.01,0.08844672203063965
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,4,32,power_law_1.01,0.09165311813354492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,4,32,power_law_1.01,0.11749247550964355
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,4,32,balanced,1.4649125671386718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,4,32,power_law_1.01,0.12526592254638672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,4,32,power_law_1.01,0.13163904190063475
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,4,32,power_law_1.01,0.12878080368041994
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,4,32,power_law_1.01,0.1338956832885742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,4,32,power_law_1.01,0.14060288429260254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,4,32,power_law_1.01,0.14538751602172853
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,4,32,power_law_1.01,0.16565248489379883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,4,32,balanced,1.9268544006347657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,4,32,balanced,0.995125732421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,4,32,power_law_1.01,0.17629312515258788
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,4,32,power_law_1.01,0.19172992706298828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,4,32,power_law_1.01,0.23448320388793947
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,4,32,power_law_1.01,0.273306884765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,4,32,power_law_1.01,0.3537804794311524
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,4,32,power_law_1.01,0.46157695770263674
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,4,32,power_law_1.01,0.6701939392089844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,4,32,power_law_1.01,0.1514470386505127
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,4,32,power_law_1.01,0.17873151779174806
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,4,32,power_law_1.01,0.9213977813720703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,4,32,power_law_1.01,0.05682303905487061
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,4,32,power_law_1.01,0.057274880409240725
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,4,32,power_law_1.01,1.154324493408203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,4,32,power_law_1.01,0.05782911777496338
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,4,32,power_law_1.01,0.05962751865386963
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,4,32,power_law_1.01,0.060014081001281736
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,2,64,power_law_1.2,5.526092529296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,4,32,power_law_1.01,0.07362815856933594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,4,32,power_law_1.01,0.07575551986694336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,4,32,power_law_1.01,0.07659008026123046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,4,32,power_law_1.01,1.5595263671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,4,32,power_law_1.01,0.08674943923950196
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,4,32,power_law_1.01,0.08183808326721191
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,4,32,power_law_1.01,0.09380352020263671
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,4,32,power_law_1.01,0.09588224411010741
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,4,32,power_law_1.01,0.12093824386596679
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,4,32,power_law_1.01,0.11453184127807617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,4,32,power_law_1.01,0.11516160011291503
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,4,32,power_law_1.01,0.13213567733764647
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,4,32,power_law_1.01,0.1745996856689453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,4,32,power_law_1.01,0.19565568923950197
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,4,32,power_law_1.01,2.4954278564453127
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,4,32,power_law_1.01,0.2683174324035645
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,4,32,power_law_1.01,0.33825790405273437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,4,32,power_law_1.01,0.4867379379272461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,4,32,power_law_1.01,0.6678514862060547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,4,32,power_law_1.01,0.8730252838134765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,4,32,power_law_1.01,0.069235200881958
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,4,32,power_law_1.2,0.08465408325195313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,4,32,power_law_1.01,3.6379608154296874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,4,32,power_law_1.2,0.08574336051940919
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,4,32,power_law_1.2,0.088090877532959
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,4,32,power_law_1.01,1.313387451171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,4,32,power_law_1.2,0.09385855674743653
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,4,32,power_law_1.2,0.12401408195495604
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,4,32,power_law_1.2,0.11808511734008789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,4,32,power_law_1.2,0.1307430362701416
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,4,32,power_law_1.2,0.1267686367034912
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,4,32,power_law_1.2,0.12818559646606445
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,4,32,power_law_1.2,0.13952383995056153
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,4,32,power_law_1.2,0.1514496040344238
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,4,32,power_law_1.2,0.15296640396118164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,4,32,power_law_1.2,0.18002431869506835
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,4,32,power_law_1.01,1.8857189941406252
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,4,32,power_law_1.2,0.18427391052246095
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,4,32,power_law_1.2,0.1849113655090332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,4,32,power_law_1.2,0.1938572883605957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,4,32,power_law_1.2,0.25679744720458986
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,4,32,power_law_1.2,0.08577664375305176
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,4,32,power_law_1.2,0.31290111541748045
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,4,32,power_law_1.2,0.4487001419067383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,4,32,power_law_1.2,0.5387020874023437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,4,32,power_law_1.01,2.5987429809570313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,4,32,power_law_1.2,0.7419904327392578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,4,32,power_law_1.2,0.05661312103271484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,4,32,power_law_1.2,0.0611353588104248
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,4,32,power_law_1.2,1.022323226928711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,4,32,power_law_1.2,0.05806464195251465
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,4,32,power_law_1.2,0.059921917915344236
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,4,32,power_law_1.2,0.059208960533142084
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,4,32,power_law_1.2,0.07038591861724854
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,4,32,power_law_1.2,0.07498112201690674
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,4,32,power_law_1.2,0.07747200012207031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,4,32,power_law_1.2,0.07529727935791015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,4,32,power_law_1.2,1.6436466979980469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,4,32,power_law_1.2,0.07974656105041504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,4,32,power_law_1.2,0.08729984283447265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,4,32,power_law_1.2,0.10581888198852538
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,4,32,power_law_1.2,0.11840383529663086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,4,32,power_law_1.2,1.9981120300292967
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,4,32,power_law_1.2,0.11716223716735841
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,4,32,power_law_1.2,0.13613823890686033
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,4,32,power_law_1.2,0.19162111282348632
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,4,32,power_law_1.2,0.23944320678710937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,4,32,power_law_1.2,0.3225702285766602
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,4,32,power_law_1.2,0.4226188659667969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,4,32,power_law_1.2,0.09626367568969726
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,4,32,power_law_1.2,0.5795443344116211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,4,32,power_law_1.2,0.12290944099426268
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,4,32,power_law_1.2,3.04455810546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,4,32,power_law_1.2,0.7145279693603516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,8,16,balanced,0.025096960067749023
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,8,16,balanced,0.03877887964248657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,8,16,balanced,0.048095998764038084
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,8,16,balanced,0.06289919853210449
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,8,16,balanced,0.08660608291625976
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,8,16,balanced,0.10187007904052733
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,8,16,balanced,0.12895615577697755
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,8,16,balanced,0.13143039703369142
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,8,16,balanced,0.140250883102417
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,8,16,balanced,0.14325119972229003
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,8,16,balanced,0.14837247848510743
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,8,16,balanced,0.1601907157897949
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,8,16,balanced,0.16542720794677734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,8,16,balanced,0.18172544479370117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,8,16,balanced,0.16823551177978516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,8,16,balanced,0.1797644805908203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,4,32,power_law_1.2,1.549160919189453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,8,16,balanced,0.19757823944091796
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,8,16,balanced,0.2357414436340332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,8,16,balanced,0.2723788833618164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,4,32,power_law_1.2,1.087479019165039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,8,16,balanced,0.37366912841796873
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,8,16,balanced,0.4552614212036133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,8,16,balanced,0.6527833557128906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,4,32,power_law_1.2,2.2855909729003905
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,8,16,balanced,0.026152958869934084
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,8,16,balanced,0.02766592025756836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,4,32,power_law_1.2,2.878367919921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,8,16,balanced,0.03263360023498535
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,8,16,balanced,1.2100224304199219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,8,16,balanced,0.04523136138916016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,8,16,balanced,0.06256768226623535
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,8,16,balanced,0.07495296001434326
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,8,16,balanced,1.5462527465820313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,8,16,balanced,0.08379648208618165
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,8,16,balanced,0.08556032180786133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,8,16,balanced,0.09107968330383301
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,8,16,balanced,0.8078221130371095
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,8,16,balanced,0.1009932804107666
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,8,16,balanced,0.11257984161376952
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,8,16,balanced,0.12073087692260742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,8,16,balanced,0.13434623718261718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,8,16,balanced,0.11946368217468262
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,8,16,balanced,0.1315392017364502
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,8,16,balanced,0.1482585620880127
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,8,16,balanced,0.1875315284729004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,8,16,balanced,0.22427263259887695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,8,16,balanced,0.2970201683044434
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,8,16,balanced,0.3695142364501953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,8,16,balanced,0.09528063774108887
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,8,16,balanced,0.5509529495239258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,8,16,balanced,3.0079141235351563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,8,16,balanced,0.6939481353759765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,8,16,balanced,2.292627258300781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,8,16,power_law_1.01,0.06913407802581786
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,8,16,balanced,1.0280115509033203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,8,16,power_law_1.01,0.08394240379333497
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,8,16,power_law_1.01,0.0882316780090332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,8,16,balanced,1.3210662841796874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,8,16,power_law_1.01,0.0934438419342041
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,8,16,power_law_1.01,0.09315839767456055
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,4,32,power_law_1.2,4.106900329589844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,8,16,power_law_1.01,0.12801024436950684
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,8,16,power_law_1.01,0.12290559768676759
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,8,16,power_law_1.01,0.12437376022338867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,8,16,power_law_1.01,0.13446399688720703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,8,16,power_law_1.01,0.1508467197418213
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,8,16,power_law_1.01,0.1493222427368164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,8,16,power_law_1.01,0.15718400001525878
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,8,16,power_law_1.01,0.1763212776184082
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,8,16,power_law_1.01,0.16058752059936524
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,8,16,balanced,1.9529370117187501
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,8,16,power_law_1.01,0.1664793586730957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,8,16,power_law_1.01,0.18967168807983398
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,8,16,power_law_1.01,0.23347711563110352
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,8,16,power_law_1.01,0.29149824142456054
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,8,16,power_law_1.01,0.39390590667724606
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,8,16,balanced,2.5693823242187497
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,8,16,power_law_1.01,0.4636902236938477
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,8,16,power_law_1.01,0.6973260498046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,8,16,power_law_1.01,0.04982016086578369
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,8,16,power_law_1.01,0.08930560111999511
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,8,16,power_law_1.01,0.9102182769775391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,8,16,power_law_1.01,0.060729598999023436
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,8,16,power_law_1.01,0.06099199771881103
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,8,16,power_law_1.01,0.06358911991119384
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,8,16,power_law_1.01,1.304565734863281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,8,16,power_law_1.01,0.06761856079101562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,8,16,power_law_1.01,0.08860544204711915
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,8,16,power_law_1.01,0.07166336059570313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,8,16,power_law_1.01,0.07879551887512207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,8,16,power_law_1.01,0.07966464042663575
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,8,16,power_law_1.01,0.09383551597595215
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,8,16,power_law_1.01,0.10361599922180176
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,8,16,power_law_1.01,0.11154303550720215
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,8,16,power_law_1.01,0.12199423789978028
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,8,16,power_law_1.01,0.1126579189300537
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,8,16,power_law_1.01,0.13710847854614258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,8,16,power_law_1.01,0.17624191284179686
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,8,16,power_law_1.01,0.2051968002319336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,8,16,power_law_1.01,2.4317388916015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,8,16,power_law_1.01,0.29391231536865237
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,8,16,power_law_1.01,0.08697855949401856
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,8,16,power_law_1.01,0.37037567138671873
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,8,16,power_law_1.01,1.6454246520996094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,8,16,power_law_1.01,0.5279487991333008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,8,16,power_law_1.01,0.11811840057373046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,8,16,power_law_1.2,0.06933760166168212
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,8,16,power_law_1.01,0.721871337890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,8,16,power_law_1.2,0.08322560310363769
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,8,16,power_law_1.01,3.42337158203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,8,16,power_law_1.01,1.0097331237792968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,8,16,power_law_1.2,0.08451456069946289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,8,16,power_law_1.2,0.08851584434509277
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,8,16,power_law_1.2,0.08794879913330078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,8,16,power_law_1.2,0.12200063705444335
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,8,16,power_law_1.2,0.10040448188781739
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,8,16,power_law_1.2,0.12966527938842773
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,8,16,power_law_1.2,0.12925439834594726
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,8,16,power_law_1.01,1.284209899902344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,8,16,power_law_1.2,0.13460351943969726
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,8,16,power_law_1.2,0.1424563217163086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,8,16,power_law_1.2,0.18012672424316406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,8,16,power_law_1.2,0.17721855163574218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,8,16,power_law_1.2,0.16390272140502932
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,8,16,power_law_1.2,0.19117696762084962
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,8,16,power_law_1.2,0.24022016525268555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,8,16,power_law_1.2,0.2928486442565918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,8,16,power_law_1.2,0.3908262252807617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,8,16,power_law_1.2,0.5086259078979493
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,8,16,power_law_1.01,2.562378234863281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,8,16,power_law_1.2,0.1561676788330078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,8,16,power_law_1.2,0.1600204849243164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,8,16,power_law_1.2,0.6916966247558594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,8,16,power_law_1.2,0.049428482055664066
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,8,16,power_law_1.01,1.9768818664550782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,8,16,power_law_1.2,0.9352114868164062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,8,16,power_law_1.2,0.06021376132965088
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,8,16,power_law_1.2,0.06463232040405273
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,8,16,power_law_1.2,0.06404223918914795
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,8,16,power_law_1.2,0.0669593620300293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,8,16,power_law_1.2,0.06904831886291504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,8,16,power_law_1.2,0.08621824264526368
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,8,16,power_law_1.2,1.4060044860839844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,8,16,power_law_1.2,0.07823872089385986
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,8,16,power_law_1.2,0.08516736030578613
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,8,16,power_law_1.2,0.08791168212890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,8,16,power_law_1.2,0.09551103591918945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,8,16,power_law_1.2,0.10431615829467773
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,8,16,power_law_1.2,1.831083526611328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,8,16,power_law_1.2,0.11181568145751952
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,8,16,power_law_1.2,0.12848896026611328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,8,16,power_law_1.2,0.11310976028442384
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,8,16,power_law_1.2,0.12407039642333983
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,8,16,power_law_1.2,0.1403238391876221
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,8,16,power_law_1.2,0.1848806381225586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,8,16,power_law_1.2,0.2338252830505371
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,8,16,power_law_1.2,0.31094272613525387
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,8,16,power_law_1.2,0.40011775970458985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,8,16,power_law_1.2,0.5718310546875001
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,16,8,balanced,0.026684160232543948
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,8,16,power_law_1.2,0.6885874938964844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,16,8,balanced,0.042789120674133305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,8,16,power_law_1.2,2.7778726196289063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,16,8,balanced,0.08325504302978516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,8,16,power_law_1.2,1.0524031829833984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,16,8,balanced,0.12063232421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,16,8,balanced,0.15511039733886717
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,16,8,balanced,0.16697216033935547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,16,8,balanced,0.2028223991394043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,8,16,power_law_1.2,1.4055769348144531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,16,8,balanced,0.2046976089477539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,16,8,balanced,0.20958976745605468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,16,8,balanced,0.22201343536376955
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,16,8,balanced,0.22774911880493165
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,16,8,balanced,0.24284671783447265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,16,8,balanced,0.2298918342590332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,16,8,balanced,0.23906688690185546
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,16,8,balanced,0.2597747230529785
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,16,8,balanced,0.30139007568359377
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,16,8,balanced,0.17145599365234374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,16,8,balanced,0.34285057067871094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,16,8,balanced,0.5202624130249023
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,16,8,balanced,0.6074316787719727
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,16,8,balanced,0.05906432151794434
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,8,16,power_law_1.2,2.800816650390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,16,8,balanced,0.8929446411132812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,8,16,power_law_1.2,2.0227507019042967
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,16,8,balanced,1.0440013122558593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,16,8,balanced,0.1150438404083252
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,8,16,power_law_1.2,3.5886782836914066
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,16,8,balanced,0.21391231536865235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,16,8,balanced,0.6388723373413085
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,16,8,balanced,1.5702528381347656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,16,8,balanced,1.0388159942626953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,16,8,balanced,2.003251190185547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,16,8,balanced,1.383389434814453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,16,8,balanced,1.3662284851074218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,16,8,balanced,1.3623487854003906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,16,8,balanced,2.953032836914063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,16,8,balanced,1.116346893310547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,16,8,balanced,1.1199539184570313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,16,8,balanced,1.1169395446777344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,16,8,balanced,1.1300863647460937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,16,8,balanced,1.1448281860351561
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,16,8,balanced,3.7877517700195313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,16,8,balanced,1.143240966796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,16,8,balanced,1.1563597106933594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,16,8,balanced,1.1588441467285155
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,16,8,balanced,1.178581771850586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,16,8,balanced,1.2258560180664062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,16,8,balanced,1.2826675415039062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,16,8,balanced,1.376416015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,16,8,balanced,1.4761074829101564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,16,8,power_law_1.01,0.07157504081726074
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,16,8,power_law_1.01,0.10814720153808595
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,16,8,balanced,2.6643942260742186
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,16,8,power_law_1.01,0.09531519889831544
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,16,8,balanced,2.885509033203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,16,8,power_law_1.01,0.11948415756225586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,16,8,power_law_1.01,0.1254246425628662
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,16,8,power_law_1.01,0.14511743545532227
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,16,8,power_law_1.01,0.1512883186340332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,16,8,power_law_1.01,0.1881497573852539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,16,8,power_law_1.01,0.19012863159179688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,16,8,balanced,4.252503051757812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,16,8,power_law_1.01,0.2031065559387207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,16,8,power_law_1.01,0.23071615219116212
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,16,8,balanced,4.651639099121094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,16,8,power_law_1.01,0.1474048042297363
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,16,8,power_law_1.01,0.21223680496215822
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,16,8,power_law_1.01,0.23340160369873048
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,16,8,power_law_1.01,0.18945152282714844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,16,8,power_law_1.01,0.2620236778259277
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,16,8,power_law_1.01,0.33431167602539064
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,16,8,power_law_1.01,0.395599365234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,16,8,power_law_1.01,0.5194726562499999
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,16,8,power_law_1.01,0.633072624206543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,16,8,power_law_1.01,0.9065254211425782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,16,8,balanced,7.474997558593751
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,16,8,power_law_1.01,1.1427212524414063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,16,8,power_law_1.01,0.21856767654418946
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,16,8,power_law_1.01,0.5173657608032227
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,16,8,power_law_1.01,1.6775782775878905
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,16,8,power_law_1.01,0.9659929656982422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,16,8,balanced,9.228915405273437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,16,8,balanced,0.3321152114868164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,16,8,power_law_1.01,2.2173414611816407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,16,8,power_law_1.01,0.9923289489746093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,16,8,power_law_1.01,1.1886284637451172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,16,8,power_law_1.01,3.1684185791015627
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,16,8,power_law_1.01,1.1910476684570312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,16,8,power_law_1.01,0.9754994964599609
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,16,8,power_law_1.01,4.322879943847656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,16,8,power_law_1.01,1.0120550537109376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,16,8,power_law_1.01,1.0135820770263673
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,16,8,power_law_1.01,1.0576857757568359
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,16,8,power_law_1.01,1.0693170928955078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,16,8,power_law_1.01,1.115252456665039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,16,8,power_law_1.01,1.3184614562988282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,16,8,power_law_1.01,1.4367295837402343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,16,8,power_law_1.01,1.8120921325683592
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,16,8,power_law_1.01,2.101326141357422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,16,8,power_law_1.01,2.7731494140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,16,8,power_law_1.01,3.401422119140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,16,8,power_law_1.01,0.7990988922119141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,16,8,power_law_1.01,1.2961946105957032
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,16,8,power_law_1.01,4.804908752441406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,16,8,power_law_1.01,1.1459801483154297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,16,8,power_law_1.2,0.07164159774780274
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,16,8,power_law_1.2,0.08716287612915039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,16,8,power_law_1.2,0.11096575736999512
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,16,8,power_law_1.2,0.1205414390563965
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,16,8,power_law_1.2,0.14264703750610352
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,16,8,power_law_1.2,0.14381055831909179
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,16,8,power_law_1.2,0.1520038414001465
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,16,8,power_law_1.01,6.0910986328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,16,8,power_law_1.2,0.18217727661132813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,16,8,power_law_1.2,0.09657088279724121
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,16,8,power_law_1.2,0.19302400588989257
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,16,8,power_law_1.2,0.20892160415649413
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,16,8,power_law_1.2,0.215231990814209
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,16,8,power_law_1.01,0.9704307556152344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,16,8,power_law_1.2,0.22974079132080077
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,16,8,power_law_1.2,0.250773754119873
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,16,8,power_law_1.2,0.27808256149291993
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,16,8,power_law_1.01,0.9724147033691406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,16,8,power_law_1.2,0.34784000396728515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,16,8,power_law_1.2,0.41178112030029296
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,16,8,power_law_1.2,0.23933952331542968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,16,8,power_law_1.01,1.0841804504394532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,16,8,power_law_1.2,0.5432588958740234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,16,8,power_law_1.2,0.6474956512451172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,16,8,power_law_1.2,0.18706432342529297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,16,8,power_law_1.2,0.9308006286621093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,16,8,power_law_1.2,1.1919026947021485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,16,8,power_law_1.2,0.5161254501342774
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,16,8,power_law_1.2,1.7217510986328126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,16,8,power_law_1.01,9.092879638671874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,16,8,power_law_1.2,0.7305741119384765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,16,8,power_law_1.2,0.7985881805419922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,16,8,power_law_1.2,2.2460557556152345
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,16,8,power_law_1.2,0.95285888671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,16,8,power_law_1.2,1.1213273620605468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,16,8,power_law_1.2,1.2485708618164062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,16,8,power_law_1.2,1.1290751647949218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,16,8,power_law_1.2,0.9535308837890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,16,8,power_law_1.2,3.364873046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,16,8,power_law_1.2,0.9564275360107422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,16,8,power_law_1.2,0.9944537353515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,16,8,power_law_1.2,1.0867123413085937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,16,8,power_law_1.01,11.574168701171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,16,8,power_law_1.2,1.1130303955078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,16,8,power_law_1.2,1.1195033264160155
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,16,8,power_law_1.2,1.168885726928711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,16,8,power_law_1.2,1.28453369140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,16,8,power_law_1.2,1.5814643859863282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,16,8,power_law_1.2,1.908143310546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,16,8,power_law_1.2,2.2986419677734373
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,16,8,power_law_1.2,3.028203430175781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,16,8,power_law_1.2,4.242517700195313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,16,8,power_law_1.2,3.842021179199219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,32,4,balanced,0.02982016086578369
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,32,4,balanced,0.04533631801605224
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,32,4,balanced,0.06837759971618652
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,32,4,balanced,0.1017740821838379
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,32,4,balanced,0.1559654426574707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,32,4,balanced,0.20876928329467775
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,16,8,power_law_1.2,4.986876220703126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,32,4,balanced,0.21713151931762695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,32,4,balanced,0.2208639907836914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,32,4,balanced,0.23221248626708985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,32,4,balanced,0.23699583053588866
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,32,4,balanced,0.24671615600585936
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,32,4,balanced,0.24327552795410154
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,16,8,power_law_1.2,6.788021850585937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,32,4,balanced,0.26516864776611326
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,32,4,balanced,0.24930559158325197
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,32,4,balanced,0.2579123115539551
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,32,4,balanced,0.2739507293701172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,32,4,balanced,0.24851200103759769
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,32,4,balanced,0.3116966438293457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,32,4,balanced,0.3502016067504883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,32,4,balanced,0.5753497695922851
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,32,4,balanced,0.6478079986572266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,16,8,power_law_1.2,1.1585727691650392
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,32,4,balanced,0.9439411163330078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,32,4,balanced,1.0953075408935546
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,32,4,balanced,1.6475622558593748
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,32,4,balanced,2.104938201904297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,32,4,balanced,3.0976766967773437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,32,4,balanced,0.34403457641601565
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,16,8,power_law_1.2,1.0325081634521485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,32,4,balanced,0.634260482788086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,32,4,balanced,3.9493133544921877
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,16,8,power_law_1.2,10.279615478515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,32,4,balanced,1.1394342041015624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,32,4,balanced,1.6206629943847655
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,32,4,balanced,1.5773388671875002
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,16,8,power_law_1.2,1.3878437805175783
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,32,4,balanced,0.12483200073242187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,32,4,balanced,1.6015884399414062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,32,4,balanced,1.5855538940429688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,32,4,balanced,0.21142784118652344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,32,4,balanced,1.5988531494140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,32,4,balanced,1.5947123718261718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,32,4,balanced,1.5241920471191406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,32,4,balanced,1.5317120361328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,32,4,balanced,1.5837887573242189
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,32,4,balanced,1.5339840698242189
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,32,4,balanced,1.5692863464355469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,32,4,balanced,1.592841033935547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,32,4,balanced,1.6398284912109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,32,4,balanced,1.6818572998046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,32,4,balanced,1.7863641357421876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,32,4,balanced,1.9096832275390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,16,8,power_law_1.2,11.55909423828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,32,4,balanced,3.442074890136719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,32,4,power_law_1.01,0.091778564453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,32,4,power_law_1.01,0.060083198547363284
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,32,4,balanced,3.736001281738281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,32,4,power_law_1.01,0.11608320236206056
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,32,4,power_law_1.01,0.13558015823364258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,32,4,power_law_1.01,0.18504831314086914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,32,4,power_law_1.01,0.1962841606140137
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,32,4,power_law_1.01,0.19235584259033203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,32,4,power_law_1.01,0.2038310432434082
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,32,4,power_law_1.01,0.21521663665771484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,32,4,power_law_1.01,0.23791488647460937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,32,4,power_law_1.01,0.0934284782409668
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,32,4,power_law_1.01,0.2217523193359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,32,4,power_law_1.01,0.22951679229736327
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,32,4,power_law_1.01,0.24637695312500002
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,32,4,power_law_1.01,0.23812480926513674
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,32,4,power_law_1.01,0.2814092826843262
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,32,4,power_law_1.01,0.34183807373046876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,32,4,balanced,5.921879272460937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,32,4,power_law_1.01,0.43374977111816404
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,32,4,power_law_1.01,0.5576639938354492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,32,4,power_law_1.01,0.6701299285888671
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,32,4,balanced,5.4269171142578125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,32,4,power_law_1.01,0.9416563415527344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,32,4,power_law_1.01,1.1619865417480468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,32,4,power_law_1.01,0.3247782516479492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,32,4,power_law_1.01,0.2534860801696778
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,32,4,power_law_1.01,1.7050253295898437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,32,4,power_law_1.01,0.5829644775390624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,32,4,power_law_1.01,0.5995072174072266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,32,4,power_law_1.01,2.2396441650390626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,32,4,power_law_1.01,0.7937010955810547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,32,4,balanced,9.360582275390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,32,4,power_law_1.01,1.0336038208007812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,32,4,power_law_1.01,3.3117428588867184
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,32,4,power_law_1.01,1.2877017211914061
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,32,4,power_law_1.01,1.5250253295898437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,32,4,power_law_1.01,1.3565362548828124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,32,4,power_law_1.01,4.243281860351563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,32,4,power_law_1.01,1.4157069396972655
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,32,4,power_law_1.01,1.38084228515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,32,4,power_law_1.01,1.3750566101074218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,32,4,power_law_1.01,1.3729446411132813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,32,4,balanced,11.758887939453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,32,4,power_law_1.01,1.4100531005859376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,32,4,power_law_1.01,1.4306343078613282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,32,4,power_law_1.01,1.4763687133789063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,32,4,power_law_1.01,1.7384921264648436
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,32,4,power_law_1.01,1.86974853515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,32,4,power_law_1.01,2.2452543640136717
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,32,4,power_law_1.01,2.6073394775390626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,32,4,power_law_1.01,1.367703094482422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,32,4,power_law_1.01,3.331261291503906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,32,4,power_law_1.2,0.05980927944183349
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,32,4,power_law_1.2,0.07649663925170899
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,32,4,power_law_1.01,4.156058959960937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,32,4,power_law_1.2,0.09138048171997071
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,32,4,power_law_1.2,0.12201215744018554
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,32,4,power_law_1.2,0.12873215675354005
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,32,4,power_law_1.2,0.1825062370300293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,32,4,power_law_1.2,0.18150527954101564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,32,4,power_law_1.2,0.19368192672729492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,32,4,power_law_1.2,0.20649599075317382
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,32,4,power_law_1.2,0.220577278137207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,32,4,power_law_1.2,0.24169984817504883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,32,4,power_law_1.2,0.22544767379760744
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,32,4,power_law_1.2,0.2335296058654785
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,32,4,power_law_1.01,5.62656005859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,32,4,power_law_1.2,0.25803136825561523
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,32,4,power_law_1.2,0.2498969650268555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,32,4,power_law_1.2,0.2652032089233399
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,32,4,power_law_1.2,0.29668479919433594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,32,4,power_law_1.2,0.3676172637939453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,32,4,power_law_1.2,0.4517542266845703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,32,4,power_law_1.2,0.5651481628417969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,32,4,power_law_1.2,0.6788582611083984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,32,4,power_law_1.2,0.9567743682861328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,32,4,power_law_1.01,7.15917724609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,32,4,power_law_1.2,1.1910694122314454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,32,4,power_law_1.2,0.32783615112304687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,32,4,power_law_1.2,1.7735513305664063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,32,4,power_law_1.2,0.4488076782226562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,32,4,power_law_1.2,0.5879743957519532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,32,4,power_law_1.2,2.3398016357421874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,32,4,power_law_1.01,1.4494808959960939
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,32,4,power_law_1.2,0.8402521514892578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,32,4,power_law_1.2,0.9603699493408204
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,32,4,power_law_1.2,1.3269850158691407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,32,4,power_law_1.2,1.374757080078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,32,4,power_law_1.2,3.349099426269531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,32,4,power_law_1.2,1.2888677978515626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,32,4,power_law_1.2,1.3499609375000001
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,32,4,power_law_1.01,10.579700927734375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,32,4,power_law_1.2,1.4105958557128906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,32,4,power_law_1.2,4.5996069335937495
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,32,4,power_law_1.2,1.421395263671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,32,4,power_law_1.2,1.346045379638672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,32,4,power_law_1.2,1.4199436950683595
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,32,4,power_law_1.2,1.5073753356933595
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,32,4,power_law_1.2,1.4947212219238282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,32,4,power_law_1.2,1.5500274658203126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,32,4,power_law_1.2,1.6918566894531248
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,32,4,power_law_1.2,1.8545318603515626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,32,4,power_law_1.2,2.0055949401855466
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,32,4,power_law_1.2,2.4175654602050782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,32,4,power_law_1.01,13.449290771484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,32,8,balanced,0.02206592082977295
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,32,8,balanced,0.021968638896942137
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,32,8,balanced,0.021882879734039306
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,32,8,balanced,0.034146559238433835
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,32,8,balanced,0.03427455902099609
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,32,8,balanced,0.0347379207611084
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,32,8,balanced,0.035143680572509765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,32,8,balanced,0.03531264066696167
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,32,8,balanced,0.03578880071640015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,32,8,balanced,0.03632384061813355
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,32,8,balanced,0.0364736008644104
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,32,4,power_law_1.2,2.80875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,32,8,balanced,0.03719295978546143
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,32,8,balanced,0.039121921062469485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,32,8,balanced,0.040262398719787595
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,32,8,balanced,0.04013055801391602
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,32,8,balanced,0.04195072174072266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,32,8,balanced,0.0433625602722168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,32,4,power_law_1.2,4.591425170898438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,32,8,balanced,0.044893441200256345
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,32,4,power_law_1.2,3.571703186035156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,32,8,balanced,0.07143040180206299
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,32,8,balanced,0.06865920066833496
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,32,8,balanced,0.08082176208496093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,32,8,balanced,0.10860159873962402
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,32,8,balanced,0.12551039695739746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,32,8,balanced,0.19763584136962892
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,32,8,balanced,0.023249919414520266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,32,8,balanced,0.2527449607849121
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,32,8,balanced,0.02335103988647461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,32,8,balanced,0.023230719566345214
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,32,8,balanced,0.03934592008590698
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,32,8,balanced,0.03934463977813721
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,32,8,balanced,0.03953151941299439
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,32,8,balanced,0.03964416027069092
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,32,4,power_law_1.2,5.932315063476563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,32,8,balanced,0.09051648139953614
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,32,8,balanced,0.037658879756927485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,32,8,balanced,0.04017536163330078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,32,8,balanced,0.04020991802215576
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,32,8,balanced,0.040181760787963865
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,32,8,balanced,0.04070911884307861
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,32,8,balanced,0.041103358268737795
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,32,8,balanced,0.04156928062438965
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,32,8,balanced,0.04208384037017822
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,32,8,balanced,0.042607359886169434
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,32,8,balanced,0.045793280601501465
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,32,8,balanced,0.04775680065155029
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,32,8,balanced,0.05609856128692627
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,32,8,balanced,0.05087103843688965
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,32,8,balanced,0.057921280860900884
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,32,8,balanced,0.06782336235046386
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,32,8,balanced,0.07890304088592529
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,32,8,balanced,0.03984512090682983
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,32,8,balanced,0.09395711898803712
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,32,8,balanced,0.12558079719543458
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,32,8,power_law_1.01,0.02199552059173584
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,32,8,balanced,0.15352831840515138
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,32,8,power_law_1.01,0.033701119422912595
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,32,8,power_law_1.01,0.03392384052276611
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,32,8,balanced,0.04419328212738037
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,32,8,power_law_1.01,0.03427072048187256
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,32,8,power_law_1.01,0.0346777606010437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,32,8,power_law_1.01,0.03573632001876831
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,32,4,power_law_1.2,7.663079833984375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,32,8,power_law_1.01,0.03690239906311035
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,32,8,power_law_1.01,0.037931520938873295
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,32,8,power_law_1.01,0.03849728107452392
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,32,8,power_law_1.01,0.039575040340423584
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,32,8,power_law_1.01,0.042225918769836425
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,32,8,power_law_1.01,0.043392000198364256
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,32,8,power_law_1.01,0.04459392070770264
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,32,8,power_law_1.01,0.06566271781921387
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,32,8,power_law_1.01,0.07347968101501465
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,32,8,power_law_1.01,0.08871040344238282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,32,8,power_law_1.01,0.041057281494140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,32,8,power_law_1.01,0.1055116844177246
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,32,8,power_law_1.01,0.05091072082519531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,32,8,power_law_1.01,0.07848832130432129
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,32,8,power_law_1.01,0.22686975479125976
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,32,8,power_law_1.01,0.12037887573242187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,32,8,power_law_1.01,0.0232806396484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,32,8,power_law_1.01,0.17332351684570313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,32,8,power_law_1.01,0.03933311939239502
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,32,8,power_law_1.01,0.03955071926116943
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,32,4,power_law_1.2,11.195010986328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,32,8,power_law_1.01,0.31288192749023436
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,32,4,power_law_1.2,14.98004150390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,32,8,power_law_1.01,0.4456768035888672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,32,8,power_law_1.01,0.041368322372436525
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,32,8,power_law_1.01,0.038853120803833005
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,32,8,power_law_1.01,0.04291200160980225
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,32,8,power_law_1.01,0.04405375957489014
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,32,8,power_law_1.01,0.044802560806274414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,32,8,power_law_1.01,0.03982975959777832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,32,8,power_law_1.01,0.04006783962249756
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,32,8,power_law_1.01,0.04669312000274658
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,32,8,power_law_1.01,0.04991360187530518
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,32,8,power_law_1.01,0.053913598060607915
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,32,8,power_law_1.01,0.05276544094085693
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,32,8,power_law_1.01,0.057392640113830565
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,32,8,power_law_1.01,0.06354944229125976
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,32,8,power_law_1.01,0.07385471820831299
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,32,8,power_law_1.01,0.0403059196472168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,32,8,power_law_1.01,0.0923788833618164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,32,8,power_law_1.01,0.11341567993164063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,32,8,power_law_1.01,0.040933117866516114
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,32,8,power_law_1.01,0.04192768096923828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,32,8,power_law_1.2,0.022135040760040282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,32,8,power_law_1.2,0.03363840103149414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,32,8,power_law_1.2,0.0341811203956604
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,32,8,power_law_1.2,0.03396991968154907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,32,8,power_law_1.2,0.03469183921813965
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,32,8,power_law_1.2,0.0359500789642334
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,32,8,power_law_1.2,0.03697792053222656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,32,8,power_law_1.01,0.14253824234008788
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,32,8,power_law_1.2,0.03755520105361938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,32,8,power_law_1.2,0.03941632032394409
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,32,8,power_law_1.2,0.039851520061492916
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,32,8,power_law_1.2,0.04114431858062744
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,32,8,power_law_1.2,0.0427839994430542
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,32,8,power_law_1.2,0.04358016014099121
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,32,8,power_law_1.2,0.044295678138732905
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,32,8,power_law_1.2,0.06853504180908203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,32,8,power_law_1.2,0.05131135940551758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,32,8,power_law_1.2,0.0702348804473877
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,32,8,power_law_1.01,0.22196352005004885
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,32,8,power_law_1.2,0.07655935764312745
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,32,8,power_law_1.2,0.08980735778808593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,32,8,power_law_1.01,0.8645299530029297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,32,8,power_law_1.2,0.10327551841735841
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,32,8,power_law_1.01,0.6701107025146484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,32,8,power_law_1.2,0.12484352111816406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,32,8,power_law_1.2,0.17778175354003906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,32,8,power_law_1.2,0.24677503585815427
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,32,8,power_law_1.2,0.3176934432983399
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,32,8,power_law_1.2,0.02321791887283325
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,32,8,power_law_1.2,0.038913280963897706
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,32,8,power_law_1.2,0.039512319564819334
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,32,8,power_law_1.2,0.039366400241851805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,32,8,power_law_1.2,0.04021759986877441
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,32,8,power_law_1.2,0.04055935859680176
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,32,8,power_law_1.2,0.04117248058319092
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,32,8,power_law_1.2,0.0418393611907959
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,32,8,power_law_1.01,0.2831142425537109
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,32,8,power_law_1.2,0.042329602241516114
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,32,8,power_law_1.2,0.04356991767883301
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,32,8,power_law_1.2,0.044788479804992676
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,32,8,power_law_1.2,0.04559232234954834
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,32,8,power_law_1.2,0.0475878381729126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,32,8,power_law_1.2,0.04978559970855713
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,32,8,power_law_1.2,0.05153920173645019
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,32,8,power_law_1.2,0.05143424034118652
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,32,8,power_law_1.2,0.4457356643676758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,32,8,power_law_1.2,0.05689599990844727
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,32,8,power_law_1.2,0.06494976043701171
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,32,8,power_law_1.2,0.039697918891906735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,32,8,power_law_1.2,0.07723008155822754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,32,8,power_law_1.2,0.09249792098999024
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,32,8,power_law_1.2,0.123187198638916
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,32,8,power_law_1.01,0.4127219009399414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,32,8,power_law_1.2,0.14926464080810548
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,32,8,balanced,0.027820799350738522
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,32,8,power_law_1.2,0.2262873649597168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,32,8,balanced,0.027740159034729005
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,32,8,balanced,0.028038399219512938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,32,8,balanced,0.045574398040771485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,32,8,balanced,0.04433023929595947
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,32,8,balanced,0.04506112098693847
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,32,8,balanced,0.04707712173461914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,32,8,balanced,0.04747776031494141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,32,8,power_law_1.01,0.5578905487060547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,32,8,balanced,0.04730112075805664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,32,8,balanced,0.04839680194854736
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,32,8,balanced,0.051016960144042965
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,32,8,balanced,0.05299456119537353
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,32,8,balanced,0.05453311920166015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,32,8,balanced,0.05379327774047852
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,32,8,balanced,0.05812736034393311
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,32,8,balanced,0.04810751914978027
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,32,8,balanced,0.08538751602172852
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,32,8,balanced,0.05056128025054931
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,32,8,balanced,0.10960255622863771
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,32,8,balanced,0.10650367736816406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,32,8,balanced,0.12462592124938965
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,32,8,balanced,0.1361996841430664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,32,8,balanced,0.16473855972290039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,32,8,power_law_1.2,0.26674943923950195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,32,8,balanced,0.19017087936401367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,32,8,balanced,0.3182988739013672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,32,8,balanced,0.02720896005630493
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,32,8,balanced,0.3820582580566406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,32,8,balanced,0.09202816009521483
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,32,8,balanced,0.02707711935043335
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,32,8,balanced,0.04822527885437012
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,32,8,balanced,0.048253440856933595
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,32,8,balanced,0.048565759658813476
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,32,8,balanced,0.04856319904327393
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,32,8,balanced,0.048730878829956054
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,32,8,balanced,0.048990721702575686
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,32,8,balanced,0.04891392230987549
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,32,8,balanced,0.04906879901885987
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,32,8,balanced,0.049571838378906254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,32,8,balanced,0.04981760025024414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,32,8,balanced,0.050446081161499026
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,32,8,balanced,0.051179518699646
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,32,8,balanced,0.05169919967651367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,32,8,balanced,0.02728447914123535
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,32,8,balanced,0.05348608016967773
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,32,8,balanced,0.05625728130340576
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,32,8,balanced,0.0592460823059082
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,32,8,balanced,0.07127039909362794
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,32,8,balanced,0.06557184219360351
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,32,8,balanced,0.07730944156646728
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,32,8,balanced,0.08571776390075683
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,32,8,balanced,0.10562944412231445
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,32,8,balanced,0.12707712173461913
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,32,8,balanced,0.1687820816040039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,32,8,power_law_1.01,0.027746560573577883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,32,8,power_law_1.01,0.044591360092163086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,32,8,power_law_1.01,0.045932798385620116
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,32,8,power_law_1.01,0.04577536106109619
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,32,8,power_law_1.01,0.046661119461059566
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,32,8,power_law_1.01,0.04821760177612305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,32,8,power_law_1.01,0.04995711803436279
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,32,8,power_law_1.01,0.05103360176086426
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,32,8,power_law_1.01,0.053038082122802734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,32,8,power_law_1.01,0.053291521072387694
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,32,8,power_law_1.01,0.058745598793029784
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,32,8,power_law_1.01,0.06357888221740723
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,32,8,power_law_1.01,0.08057855606079102
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,32,8,power_law_1.01,0.08790528297424316
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,32,8,power_law_1.01,0.09534848213195801
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,32,8,balanced,0.20852096557617186
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,32,8,power_law_1.01,0.11333760261535644
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,32,8,power_law_1.01,0.10598655700683594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,32,8,power_law_1.01,0.1216102409362793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,32,8,power_law_1.01,0.13300864219665526
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,32,8,power_law_1.01,0.17653247833251953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,32,8,power_law_1.2,0.6789119720458985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,32,8,power_law_1.01,0.15748991966247558
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,32,8,power_law_1.01,0.2652710342407226
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,32,8,power_law_1.01,0.4270182418823242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,32,8,power_law_1.01,0.027393279075622556
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,32,8,power_law_1.2,0.8986188507080077
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,32,8,power_law_1.01,0.04734208106994629
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,32,8,power_law_1.01,0.048380160331726076
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,32,8,power_law_1.01,0.04832255840301514
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,32,8,power_law_1.01,0.04850048065185547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,32,8,power_law_1.01,0.04890624046325683
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,32,8,power_law_1.01,0.04933375835418701
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,32,8,power_law_1.01,0.04987904071807862
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,32,8,power_law_1.01,0.05032832145690917
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,32,8,power_law_1.01,0.37496959686279296
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,32,8,power_law_1.01,0.05087488174438477
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,32,8,power_law_1.01,0.052030720710754395
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,32,8,power_law_1.01,0.05350783824920654
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,32,8,power_law_1.01,0.055125761032104495
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,32,8,power_law_1.01,0.06060031890869141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,32,8,power_law_1.01,0.06503551959991455
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,32,8,power_law_1.01,0.06950528144836425
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,32,8,power_law_1.01,0.06950784206390381
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,32,8,power_law_1.01,0.07563136100769044
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,32,8,power_law_1.01,0.08339584350585938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,32,8,power_law_1.01,0.10268671989440918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,32,8,power_law_1.01,0.19936256408691405
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,32,8,power_law_1.01,0.11438079833984376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,32,8,power_law_1.01,0.9833650970458985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,32,8,power_law_1.01,0.16891391754150392
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,32,8,power_law_1.01,0.7164313507080078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,32,8,power_law_1.2,0.02818943977355957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,32,8,power_law_1.2,0.04529280185699463
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,32,8,power_law_1.2,0.04574463844299316
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,32,8,power_law_1.2,0.045899519920349116
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,32,8,power_law_1.2,0.046647038459777836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,32,8,power_law_1.2,0.04728960037231446
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,32,8,power_law_1.2,0.049977598190307615
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,32,8,power_law_1.2,0.05148032188415528
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,32,8,power_law_1.2,0.05353983879089356
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,32,8,power_law_1.2,0.05528448104858399
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,32,8,power_law_1.01,1.3159207153320314
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,32,8,power_law_1.2,0.06328832149505616
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,32,8,power_law_1.2,0.06085760116577148
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,32,8,power_law_1.2,0.0814310359954834
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,32,8,power_law_1.2,0.08958208084106445
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,32,8,power_law_1.2,0.09560064315795899
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,32,8,power_law_1.2,0.10457728385925294
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,32,8,power_law_1.2,0.10987135887145996
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,32,8,power_law_1.2,0.1342515182495117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,32,8,power_law_1.01,0.3138201522827148
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,32,8,power_law_1.2,0.42392959594726565
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,32,8,power_law_1.2,0.1225881576538086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,32,8,power_law_1.01,0.5344460678100587
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,32,8,power_law_1.2,0.15539456367492677
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,32,8,power_law_1.01,0.37706111907958983
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,32,8,power_law_1.2,0.17732351303100585
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,32,8,power_law_1.2,0.3542860794067383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,32,8,power_law_1.2,0.02713855981826782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,32,8,power_law_1.2,0.04801536083221435
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,32,8,power_law_1.2,0.048336639404296874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,32,8,power_law_1.2,0.048336639404296874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,32,8,power_law_1.2,0.04866687774658203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,32,8,power_law_1.2,0.04896512031555176
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,32,8,power_law_1.2,0.0494271993637085
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,32,8,power_law_1.2,0.05009920120239257
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,32,8,power_law_1.2,0.05063936233520507
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,32,8,power_law_1.2,0.2965888023376465
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,32,8,power_law_1.2,0.0511027193069458
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,32,8,power_law_1.2,0.05308800220489503
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,32,8,power_law_1.2,0.6575769805908204
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,32,8,power_law_1.2,0.05383935928344726
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,32,8,power_law_1.2,0.05547776222229004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,32,8,power_law_1.2,0.05812607765197754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,32,8,power_law_1.2,0.06461952209472657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,32,8,power_law_1.2,0.06898303985595702
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,32,8,power_law_1.2,0.06681087970733643
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,32,8,power_law_1.2,0.07667712211608887
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,32,8,power_law_1.01,0.7573887634277343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,32,8,power_law_1.2,0.08441856384277344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,32,8,power_law_1.2,0.10393983840942382
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,32,8,power_law_1.2,0.11703040122985839
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,32,8,power_law_1.2,0.5496614456176758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,32,8,power_law_1.2,0.5570508956909179
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,32,8,power_law_1.2,0.1683302307128906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,32,8,power_law_1.2,0.20853120803833008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,1,256,balanced,0.06803967952728271
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,1,256,balanced,0.06823040008544921
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,1,256,balanced,0.06818560123443604
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,1,256,balanced,0.06816256046295166
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,1,256,balanced,0.06815231800079345
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,1,256,balanced,0.0682150411605835
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,1,256,balanced,0.12349056243896483
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,1,256,balanced,0.12418560028076171
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,1,256,balanced,0.12309632301330567
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,1,256,balanced,0.12355839729309082
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,1,256,balanced,0.12315520286560058
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,1,256,balanced,0.1246566390991211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,1,256,balanced,0.12386816024780274
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,1,256,balanced,0.1245030403137207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,1,256,balanced,0.12607999801635744
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,1,256,balanced,0.12720000267028808
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,1,256,balanced,0.1271129608154297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,32,8,power_law_1.2,0.35629184722900387
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,1,256,balanced,0.12974720001220702
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,1,256,balanced,0.1349516773223877
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,1,256,balanced,0.1404595184326172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,1,256,balanced,0.14356351852416993
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,1,256,balanced,0.14794495582580566
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,1,256,balanced,0.1457907199859619
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,1,256,balanced,0.15836288452148437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,1,256,balanced,0.16959743499755858
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,1,256,balanced,0.20634624481201175
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,1,256,balanced,0.03856127977371216
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,1,256,balanced,0.24777215957641602
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,1,256,balanced,0.03818880081176758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,1,256,balanced,0.03875328063964843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,1,256,balanced,0.03847935914993286
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,1,256,balanced,0.037740800380706784
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,1,256,balanced,0.06092415809631348
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,1,256,balanced,0.06079103946685791
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,1,256,balanced,0.061954560279846194
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,1,256,balanced,0.0606169605255127
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,1,256,balanced,0.03650176048278809
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,1,256,balanced,0.060902400016784666
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,1,256,balanced,0.061584639549255374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,1,256,balanced,0.06277247905731201
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,1,256,balanced,0.06251264095306397
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,1,256,balanced,0.06410624027252197
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,1,256,balanced,0.06406271934509278
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,1,256,balanced,0.06602367877960205
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,1,256,balanced,0.06859519958496094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,1,256,balanced,0.07088895797729491
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,32,8,power_law_1.2,0.3140928077697754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,1,256,balanced,0.07492479801177979
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,1,256,balanced,0.08165375709533691
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,1,256,balanced,0.0759500789642334
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,1,256,balanced,0.08623744010925292
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,1,256,balanced,0.09431679725646973
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,1,256,balanced,0.10943743705749512
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,1,256,balanced,0.1266329574584961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,1,256,power_law_1.01,0.0682751989364624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,1,256,power_law_1.01,0.06780159950256348
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,1,256,power_law_1.01,0.06795775890350342
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,1,256,power_law_1.01,0.06830080032348633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,1,256,balanced,0.0695577621459961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,1,256,power_law_1.01,0.12316160202026367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,1,256,power_law_1.01,0.12428288459777832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,1,256,power_law_1.01,0.12383616447448731
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,1,256,power_law_1.01,0.1246412754058838
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,1,256,power_law_1.01,0.12588671684265135
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,1,256,power_law_1.01,0.12740351676940917
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,1,256,power_law_1.01,0.1268607997894287
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,1,256,power_law_1.01,0.1294105625152588
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,1,256,power_law_1.01,0.13446656227111817
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,1,256,power_law_1.01,0.13557632446289064
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,1,256,power_law_1.01,0.14040191650390627
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,1,256,power_law_1.01,0.14607104301452636
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,1,256,power_law_1.01,0.1500915241241455
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,1,256,power_law_1.01,0.1617843246459961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,1,256,power_law_1.01,0.20277631759643552
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,1,256,power_law_1.01,0.2527833557128906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,1,256,power_law_1.01,0.3702656173706055
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,1,256,power_law_1.01,0.4438681411743164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,1,256,power_law_1.01,0.14208512306213378
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,1,256,power_law_1.01,0.6728537750244141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,1,256,power_law_1.01,0.039607040882110596
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,1,256,power_law_1.01,0.9052838134765626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,1,256,power_law_1.01,0.03847935914993286
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,1,256,power_law_1.01,0.03837055921554565
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,1,256,power_law_1.01,0.039142398834228515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,1,256,power_law_1.01,0.06230912208557129
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,1,256,power_law_1.01,1.4749658203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,1,256,power_law_1.01,0.06295040130615234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,1,256,power_law_1.01,0.0635532808303833
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,1,256,power_law_1.01,0.06250495910644531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,1,256,power_law_1.01,0.06517375946044922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,1,256,power_law_1.01,0.06499711990356445
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,1,256,power_law_1.01,0.06656640052795411
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,1,256,power_law_1.01,0.06916863918304443
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,1,256,power_law_1.01,2.0162611389160157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,1,256,power_law_1.01,0.0685260820388794
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,1,256,power_law_1.01,0.07043712139129639
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,1,256,power_law_1.01,0.07276159763336182
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,1,256,power_law_1.01,0.07537024021148682
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,1,256,power_law_1.01,0.08169471740722656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,1,256,power_law_1.01,0.08277888298034668
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,1,256,power_law_1.01,0.09000448226928712
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,32,8,power_law_1.2,1.0924646759033203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,1,256,power_law_1.01,0.10252672195434571
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,1,256,power_law_1.01,0.11984383583068847
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,1,256,power_law_1.01,0.15965439796447753
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,1,256,power_law_1.01,0.21843584060668944
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,1,256,power_law_1.01,0.2953561592102051
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,1,256,power_law_1.01,0.4080038452148438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,1,256,power_law_1.2,0.06780543804168701
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,1,256,power_law_1.01,0.6275673675537109
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,1,256,power_law_1.2,0.06868607997894287
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,1,256,power_law_1.01,0.7563404846191406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,32,8,power_law_1.2,0.5909183883666993
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,1,256,power_law_1.2,0.12328960418701171
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,1,256,power_law_1.2,0.12392576217651367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,1,256,power_law_1.2,0.12388863563537597
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,1,256,power_law_1.2,0.1239680004119873
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,1,256,power_law_1.2,0.12697983741760255
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,1,256,power_law_1.2,0.1290611171722412
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,1,256,power_law_1.2,0.1281228828430176
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,1,256,power_law_1.2,0.12964096069335937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,1,256,power_law_1.2,0.13397248268127443
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,1,256,power_law_1.2,0.134334716796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,1,256,power_law_1.2,0.06845823764801026
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,1,256,power_law_1.2,0.14131584167480468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,1,256,power_law_1.2,0.14277119636535646
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,32,8,power_law_1.2,1.4481893920898439
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,1,256,power_law_1.2,0.14595968246459962
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,1,256,power_law_1.2,0.15192319869995116
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,1,256,power_law_1.2,0.16685823440551756
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,1,256,power_law_1.2,0.20027391433715822
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,1,256,power_law_1.2,0.28172672271728516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,1,256,power_law_1.2,0.3834431838989258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,1,256,power_law_1.2,0.5881881713867188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,1,256,power_law_1.2,0.038858239650726316
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,1,256,power_law_1.2,0.8855526733398438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,1,256,power_law_1.2,0.03851392030715942
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,1,256,power_law_1.2,0.03880959987640381
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,1,256,power_law_1.2,1.2786303710937499
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,1,256,power_law_1.2,0.06082943916320801
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,1,256,power_law_1.2,0.061527037620544435
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,1,256,power_law_1.2,0.06235263824462891
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,1,256,power_law_1.2,0.06182655811309814
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,1,256,power_law_1.2,0.06613503932952881
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,1,256,power_law_1.2,1.8609703063964844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,1,256,power_law_1.2,0.06553343772888184
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,1,256,power_law_1.2,0.06453887939453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,1,256,power_law_1.2,0.07226624011993407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,1,256,power_law_1.2,0.06968832015991211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,1,256,power_law_1.2,0.07275904178619384
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,1,256,power_law_1.2,0.07843455791473389
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,1,256,power_law_1.2,0.08324095726013184
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,1,256,power_law_1.2,0.08068351745605469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,1,256,power_law_1.2,0.09181952476501465
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,1,256,power_law_1.2,0.10074496269226074
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,1,256,power_law_1.2,0.14315008163452148
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,1,256,power_law_1.2,0.16705663681030275
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,1,256,power_law_1.2,0.26325376510620113
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,1,256,power_law_1.2,0.3572159957885742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,1,256,power_law_1.2,0.06960256099700927
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,2,128,balanced,0.04865920066833496
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,2,128,balanced,0.04885119915008545
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,1,256,power_law_1.2,0.7987891387939453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,2,128,balanced,0.05012095928192138
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,1,256,power_law_1.2,3.9743988037109377
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,2,128,balanced,0.052203521728515626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,2,128,balanced,0.054195199012756344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,1,256,power_law_1.2,1.2066738891601563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,2,128,balanced,0.06969471931457519
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,2,128,balanced,0.11127039909362793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,2,128,balanced,0.1125107192993164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,2,128,balanced,0.11301247596740722
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,1,256,power_law_1.2,0.5976729583740235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,2,128,balanced,0.1147481632232666
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,2,128,balanced,0.11494144439697265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,2,128,balanced,0.11621120452880859
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,2,128,balanced,0.11887616157531737
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,1,256,power_law_1.2,1.7553996276855468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,1,256,power_law_1.2,2.8988058471679685
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,2,128,balanced,0.11910783767700195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,2,128,balanced,0.12303359985351563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,2,128,balanced,0.13608575820922852
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,2,128,balanced,0.11103487968444825
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,2,128,balanced,0.13881343841552735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,2,128,balanced,0.14860544204711915
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,2,128,balanced,0.16192384719848635
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,2,128,balanced,0.11337727546691893
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,2,128,balanced,0.1773324775695801
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,2,128,balanced,0.22608768463134768
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,2,128,balanced,0.0315827202796936
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,2,128,balanced,0.031732480525970455
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,2,128,balanced,0.27430656433105466
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,2,128,balanced,0.03146624088287354
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,2,128,balanced,0.03165184020996094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,2,128,balanced,0.1173363208770752
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,2,128,balanced,0.036058878898620604
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,2,128,balanced,0.0578380823135376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,2,128,balanced,0.057272319793701176
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,2,128,balanced,0.05705471992492676
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,2,128,balanced,0.05752064228057861
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,2,128,balanced,0.05951871871948242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,2,128,balanced,0.059650559425354
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,2,128,balanced,0.05760128021240234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,2,128,balanced,0.1296499156951904
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,2,128,balanced,0.13571840286254883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,2,128,balanced,0.060016641616821284
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,2,128,balanced,0.06546688079833984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,2,128,balanced,0.06415743827819824
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,2,128,balanced,0.06458367824554442
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,2,128,balanced,0.06971776008605957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,2,128,balanced,0.07528575897216797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,2,128,balanced,0.07061888217926025
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,2,128,balanced,0.07816703796386719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,2,128,balanced,0.08460672378540039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,2,128,balanced,0.09753727912902832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,2,128,balanced,0.11059200286865234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,2,128,balanced,0.13714943885803224
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,2,128,balanced,0.031508479118347164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,2,128,power_law_1.01,0.06334847927093505
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,2,128,power_law_1.01,0.06288640022277832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,2,128,power_law_1.01,0.06303999900817872
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,2,128,power_law_1.01,0.06315392017364502
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,2,128,power_law_1.01,0.11138431549072265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,2,128,power_law_1.01,0.11214847564697265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,2,128,power_law_1.01,0.11287039756774901
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,2,128,power_law_1.01,0.11736831665039063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,2,128,power_law_1.01,0.11592320442199706
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,2,128,balanced,0.06600575923919678
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,2,128,power_law_1.01,0.11683839797973632
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,2,128,power_law_1.01,0.11809151649475098
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,2,128,power_law_1.01,0.1209216022491455
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,2,128,power_law_1.01,0.12027520179748534
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,2,128,power_law_1.01,0.1272268772125244
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,2,128,power_law_1.01,0.13095680236816407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,2,128,power_law_1.01,0.13438336372375487
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,2,128,power_law_1.01,0.1422220802307129
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,2,128,power_law_1.01,0.15225088119506835
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,2,128,power_law_1.01,0.17171840667724608
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,2,128,power_law_1.01,0.2193280029296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,2,128,power_law_1.01,0.2766796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,2,128,power_law_1.01,0.3548761749267578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,2,128,power_law_1.01,0.5518476867675781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,2,128,power_law_1.01,0.13604864120483398
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,2,128,power_law_1.01,0.7775730895996094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,2,128,balanced,0.16551807403564453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,2,128,power_law_1.01,0.03520128011703491
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,2,128,power_law_1.01,0.036252160072326664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,2,128,power_law_1.01,0.03589632034301758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,2,128,power_law_1.01,1.0470130920410157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,2,128,power_law_1.01,0.05661695957183838
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,2,128,power_law_1.01,0.06049280166625977
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,2,128,power_law_1.01,0.059644160270690916
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,2,128,power_law_1.01,0.05873280048370362
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,2,128,power_law_1.01,0.06029695987701415
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,2,128,power_law_1.01,1.3884083557128908
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,2,128,power_law_1.01,0.061299200057983394
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,2,128,power_law_1.01,0.06636928081512451
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,2,128,power_law_1.01,0.06587520122528076
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,2,128,power_law_1.01,0.06732416152954102
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,2,128,power_law_1.01,0.035903999805450434
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,2,128,power_law_1.01,0.07228415966033935
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,2,128,power_law_1.01,0.07452928066253663
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,2,128,power_law_1.01,0.08066176414489747
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,2,128,power_law_1.01,0.08708992004394531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,2,128,power_law_1.01,0.09703680038452148
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,2,128,power_law_1.01,0.11425663948059081
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,2,128,power_law_1.01,0.15634431838989257
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,2,128,power_law_1.01,0.1981439971923828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,2,128,power_law_1.01,0.32204288482666016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,2,128,power_law_1.2,0.06256127834320069
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,2,128,power_law_1.01,0.41334270477294927
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,2,128,power_law_1.01,0.0644543981552124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,2,128,power_law_1.2,0.06329216003417969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,2,128,power_law_1.01,0.6087231826782227
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,2,128,power_law_1.01,0.07496575832366943
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,2,128,power_law_1.2,0.11060352325439453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,2,128,power_law_1.01,0.7880524444580078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,2,128,power_law_1.2,0.1107520008087158
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,2,128,power_law_1.2,0.11345408439636231
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,32,8,power_law_1.2,0.7784806060791015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,2,128,power_law_1.2,0.11598464012145997
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,2,128,power_law_1.2,0.12004351615905762
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,2,128,power_law_1.2,0.12286848068237304
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,2,128,power_law_1.2,0.13034111976623536
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,2,128,power_law_1.2,0.1349184036254883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,2,128,power_law_1.2,0.1389900779724121
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,2,128,power_law_1.2,0.13647744178771973
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,2,128,power_law_1.2,0.06273663997650146
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,2,128,power_law_1.2,0.1502015972137451
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,2,128,power_law_1.2,0.15861632347106933
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,2,128,power_law_1.2,0.22315647125244142
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,2,128,power_law_1.2,0.11337856292724609
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,2,128,power_law_1.2,0.2774771118164062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,2,128,power_law_1.2,0.11503999710083009
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,2,128,power_law_1.2,0.39350654602050783
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,2,128,power_law_1.2,0.12138496398925783
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,2,128,power_law_1.2,0.12560256004333498
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,2,128,power_law_1.2,0.4811404800415039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,2,128,power_law_1.2,0.03660928010940552
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,2,128,power_law_1.2,0.03702143907546997
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,2,128,power_law_1.2,0.03661695957183838
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,2,128,power_law_1.2,0.7800460815429687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,2,128,power_law_1.2,0.056641278266906736
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,2,128,power_law_1.2,0.06036608219146729
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,2,128,power_law_1.2,0.05841663837432861
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,2,128,power_law_1.2,0.059621119499206544
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,2,128,power_law_1.2,0.0621619176864624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,2,128,power_law_1.2,0.06642303943634034
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,2,128,power_law_1.2,0.0636032009124756
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,2,128,power_law_1.2,0.06417151927947998
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,2,128,power_law_1.2,0.06709760189056396
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,2,128,power_law_1.2,0.07212800025939942
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,2,128,power_law_1.2,0.07538815975189209
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,2,128,power_law_1.2,0.07432960033416748
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,2,128,power_law_1.2,0.07691264152526855
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,2,128,power_law_1.2,0.08603391647338868
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,2,128,power_law_1.2,1.8503514099121094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,2,128,power_law_1.2,0.09158271789550781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,2,128,power_law_1.2,0.05572864055633545
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,2,128,power_law_1.2,0.12651904106140138
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,2,128,power_law_1.2,0.1385420799255371
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,2,128,power_law_1.2,0.19603839874267578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,2,128,power_law_1.2,1.1947174072265625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,2,128,power_law_1.2,0.48180351257324217
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,4,64,balanced,0.0306060791015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,2,128,power_law_1.2,0.6208908843994141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,4,64,balanced,0.03067008018493652
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,2,128,power_law_1.2,2.8390899658203126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,4,64,balanced,0.030565121173858643
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,4,64,balanced,0.05246592044830323
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,2,128,power_law_1.2,0.9051468658447266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,4,64,balanced,0.10706560134887695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,4,64,balanced,0.10710399627685548
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,4,64,balanced,0.10767999649047852
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,4,64,balanced,0.10808575630187989
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,4,64,balanced,0.10913536071777344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,2,128,power_law_1.2,1.257017593383789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,2,128,power_law_1.2,0.2673459243774414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,4,64,balanced,0.10914048194885254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,4,64,balanced,0.1122214412689209
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,4,64,balanced,0.11132800102233886
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,4,64,balanced,0.11683967590332031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,4,64,balanced,0.12030976295471192
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,4,64,balanced,0.12257151603698731
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,4,64,balanced,0.12695808410644532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,4,64,balanced,0.12320768356323242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,4,64,balanced,0.13181695938110352
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,4,64,balanced,0.13907456398010254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,4,64,balanced,0.15180416107177735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,4,64,balanced,0.1624358367919922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,4,64,balanced,0.19016191482543945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,4,64,balanced,0.21544832229614258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,4,64,balanced,0.03058687925338745
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,4,64,balanced,0.030451200008392333
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,4,64,balanced,0.2728473663330078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,4,64,balanced,0.030327041149139405
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,4,64,balanced,0.35512958526611327
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,4,64,balanced,0.031037440299987794
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,4,64,balanced,0.03853823900222778
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,4,64,balanced,0.05912191867828369
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,4,64,balanced,0.057349119186401364
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,4,64,balanced,0.05966720104217529
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,4,64,balanced,0.05854080200195313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,4,64,balanced,0.06363647937774658
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,4,64,balanced,0.06153088092803956
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,4,64,balanced,0.0632806396484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,4,64,balanced,0.06429183959960938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,4,64,balanced,0.06390016078948975
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,4,64,balanced,0.06619135856628418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,4,64,balanced,0.06958847999572754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,4,64,balanced,0.07544447898864745
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,4,64,balanced,0.06908671855926514
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,4,64,balanced,0.07558656215667725
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,4,64,balanced,0.08114303588867187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,4,64,balanced,0.0909721565246582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,4,64,balanced,0.10201727867126464
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,4,64,balanced,0.06796927928924561
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,4,64,balanced,0.13079680442810057
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,4,64,balanced,0.15150848388671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,4,64,balanced,0.19867391586303712
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,4,64,balanced,0.03059200048446655
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,4,64,power_law_1.01,0.06173183917999268
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,4,64,balanced,0.24589439392089846
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,4,64,power_law_1.01,0.06188032150268554
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,4,64,power_law_1.01,0.06281087875366212
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,4,64,power_law_1.01,0.062126078605651856
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,4,64,power_law_1.01,0.10520575523376466
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,4,64,power_law_1.01,0.10776191711425782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,4,64,power_law_1.01,0.10925824165344238
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,4,64,power_law_1.01,0.10995583534240723
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,4,64,power_law_1.01,0.11433343887329102
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,4,64,power_law_1.01,0.11425663948059081
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,4,64,power_law_1.01,0.1192512035369873
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,4,64,balanced,0.03050112009048462
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,4,64,power_law_1.01,0.12508543968200683
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,4,64,power_law_1.01,0.1289420795440674
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,4,64,power_law_1.01,0.12902527809143066
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,4,64,power_law_1.01,0.13091456413269043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,4,64,power_law_1.01,0.1412070369720459
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,4,64,power_law_1.01,0.1501657581329346
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,4,64,power_law_1.01,0.17005184173583984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,4,64,power_law_1.01,0.19679744720458986
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,4,64,power_law_1.01,0.29367551803588865
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,4,64,power_law_1.01,0.3534873580932617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,4,64,power_law_1.01,0.4747366333007813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,4,64,power_law_1.01,0.03769216060638428
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,4,64,power_law_1.01,0.10787455558776855
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,4,64,power_law_1.01,0.03826688051223755
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,4,64,power_law_1.01,0.6279654312133789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,4,64,power_law_1.01,0.038886399269104005
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,4,64,power_law_1.01,0.038260478973388676
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,4,64,power_law_1.01,0.1221440029144287
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,4,64,power_law_1.01,0.05597184181213379
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,4,64,power_law_1.01,0.058604798316955566
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,4,64,power_law_1.01,0.8637567901611328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,4,64,power_law_1.01,0.05898496150970459
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,4,64,power_law_1.01,0.06416384220123292
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,4,64,power_law_1.01,0.0622489595413208
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,4,64,power_law_1.01,0.061861119270324706
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,4,64,power_law_1.01,0.06401408195495606
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,4,64,power_law_1.01,0.06627583980560303
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,4,64,power_law_1.01,0.07010303974151612
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,4,64,power_law_1.01,0.07457151889801025
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,4,64,power_law_1.01,1.1645235443115234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,4,64,power_law_1.01,0.07453055858612061
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,4,64,power_law_1.01,0.08457856178283692
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,4,64,power_law_1.01,0.09439616203308106
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,4,64,power_law_1.01,0.11403391838073731
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,4,64,power_law_1.01,0.13408767700195312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,4,64,power_law_1.01,0.16943487167358398
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,4,64,power_law_1.01,0.22206592559814453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,4,64,power_law_1.01,0.29733375549316404
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,4,64,power_law_1.2,0.06163328170776368
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,4,64,power_law_1.01,0.3860198211669922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,4,64,power_law_1.2,0.061835517883300786
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,4,64,power_law_1.01,0.5525260925292969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,4,64,power_law_1.2,0.06237055778503418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,4,64,power_law_1.01,0.06810880184173584
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,4,64,power_law_1.2,0.10607487678527831
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,4,64,power_law_1.01,0.07479423999786376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,4,64,power_law_1.2,0.10430080413818359
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,4,64,power_law_1.01,0.8015372467041015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,4,64,power_law_1.2,0.10730496406555176
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,4,64,power_law_1.2,0.10890111923217774
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,4,64,power_law_1.2,0.10826239585876465
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,4,64,power_law_1.2,0.11352959632873535
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,4,64,power_law_1.2,0.11085439682006837
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,4,64,power_law_1.2,0.11550848007202148
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,4,64,power_law_1.2,0.12343168258666992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,4,64,power_law_1.2,0.1277184009552002
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,4,64,power_law_1.2,0.12631936073303224
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,4,64,power_law_1.2,0.13455743789672853
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,4,64,power_law_1.2,0.15364864349365234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,4,64,power_law_1.2,0.1699724769592285
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,4,64,power_law_1.2,0.19397632598876952
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,4,64,power_law_1.2,0.2429158401489258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,4,64,power_law_1.2,0.3272422409057617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,4,64,power_law_1.2,0.42773887634277347
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,4,64,power_law_1.2,0.03789567947387695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,4,64,power_law_1.2,0.7198579406738281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,4,64,power_law_1.2,0.12148608207702635
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,4,64,power_law_1.2,0.1295206356048584
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,4,64,power_law_1.2,0.03795968055725098
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,4,64,power_law_1.2,0.8699712371826172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,4,64,power_law_1.2,0.03849728107452392
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,4,64,power_law_1.2,0.05547904014587403
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,4,64,power_law_1.2,0.05583231925964356
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,4,64,power_law_1.2,0.05921535968780518
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,4,64,power_law_1.2,0.06027647972106933
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,4,64,power_law_1.2,0.06030848026275635
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,4,64,power_law_1.2,0.06407936096191406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,4,64,power_law_1.2,0.06262656211853027
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,4,64,power_law_1.2,1.27833984375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,4,64,power_law_1.2,0.06739327907562256
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,4,64,power_law_1.2,0.06883584022521973
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,4,64,power_law_1.2,0.07657855987548828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,4,64,power_law_1.2,0.07389567852020264
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,4,64,power_law_1.2,0.07366528034210204
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,4,64,power_law_1.2,0.09188223838806152
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,4,64,power_law_1.2,0.10720383644104003
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,4,64,power_law_1.2,0.13594752311706543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,4,64,power_law_1.2,0.15776896476745605
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,4,64,power_law_1.2,1.8082943725585938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,4,64,power_law_1.2,0.23325824737548828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,4,64,power_law_1.2,0.2633062362670898
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,8,32,balanced,0.029825279712677
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,8,32,balanced,0.029945600032806396
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,4,64,power_law_1.2,0.4357324981689453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,8,32,balanced,0.029666559696197508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,4,64,power_law_1.2,0.06178559780120849
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,4,64,power_law_1.2,0.5538943862915039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,8,32,balanced,0.05430912017822266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,8,32,balanced,0.10744576454162598
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,8,32,balanced,0.10942208290100097
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,8,32,balanced,0.1074073600769043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,8,32,balanced,0.10924672126770019
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,4,64,power_law_1.2,0.9370291137695312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,8,32,balanced,0.115033597946167
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,8,32,balanced,0.11119744300842285
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,8,32,balanced,0.11629055976867675
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,8,32,balanced,0.11980799674987792
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,8,32,balanced,0.12409088134765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,8,32,balanced,0.12814975738525392
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,8,32,balanced,0.12120063781738281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,8,32,balanced,0.1270464038848877
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,8,32,balanced,0.13611519813537598
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,8,32,balanced,0.14647040367126465
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,8,32,balanced,0.15644288063049316
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,8,32,balanced,0.031123199462890626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,8,32,balanced,0.18036991119384765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,8,32,balanced,0.20020479202270508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,8,32,balanced,0.06452608108520508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,8,32,balanced,0.24641279220581053
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,8,32,balanced,0.030064640045166013
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,8,32,balanced,0.2917836761474609
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,8,32,balanced,0.029939200878143314
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,4,64,power_law_1.2,0.07749631881713867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,8,32,balanced,0.030385921001434325
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,8,32,balanced,0.032641279697418216
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,8,32,balanced,0.04175615787506103
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,8,32,balanced,0.06573311805725097
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,8,32,balanced,0.514353904724121
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,8,32,balanced,0.06267903804779053
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,8,32,balanced,0.06497024059295654
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,8,32,balanced,0.06960383892059327
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,8,32,balanced,0.06751488208770752
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,8,32,balanced,0.06762752056121826
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,8,32,balanced,0.06878208160400391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,8,32,balanced,0.0714854383468628
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,8,32,balanced,0.07460224151611328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,8,32,balanced,0.07774079799652099
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,8,32,balanced,0.07114751815795899
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,8,32,balanced,0.07715968132019044
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,8,32,balanced,0.08262399673461915
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,8,32,balanced,0.0919270420074463
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,8,32,balanced,0.10253696441650391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,8,32,balanced,0.12457344055175781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,8,32,balanced,0.14525695800781252
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,8,32,balanced,0.029920001029968262
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,8,32,balanced,0.19513599395751952
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,8,32,balanced,0.38759552001953124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,8,32,balanced,0.23350400924682618
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,8,32,power_law_1.01,0.06284160137176513
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,8,32,balanced,0.32264831542968747
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,8,32,power_law_1.01,0.06327424049377442
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,8,32,power_law_1.01,0.06382080078124999
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,8,32,power_law_1.01,0.06437759876251221
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,8,32,power_law_1.01,0.1053990364074707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,8,32,power_law_1.01,0.10829183578491211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,8,32,power_law_1.01,0.10660736083984375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,8,32,power_law_1.01,0.10990719795227051
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,8,32,power_law_1.01,0.11033344268798828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,8,32,power_law_1.01,0.11119872093200685
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,8,32,power_law_1.01,0.11609472274780272
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,8,32,power_law_1.01,0.12171135902404787
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,8,32,power_law_1.01,0.12252927780151368
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,8,32,power_law_1.01,0.12466431617736817
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,8,32,power_law_1.01,0.12751359939575196
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,8,32,power_law_1.01,0.13368191719055175
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,8,32,power_law_1.01,0.14826623916625975
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,8,32,power_law_1.01,0.15626239776611328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,8,32,power_law_1.01,0.17861888885498048
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,4,64,power_law_1.2,1.1614745330810545
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,8,32,balanced,0.4103385543823242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,8,32,power_law_1.01,0.22280960083007811
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,8,32,power_law_1.01,0.2515456008911133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,8,32,power_law_1.01,0.32065025329589847
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,8,32,power_law_1.01,0.040122880935668945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,8,32,power_law_1.01,0.3994355010986328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,8,32,power_law_1.01,0.0406822395324707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,8,32,power_law_1.01,0.1290675163269043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,8,32,power_law_1.01,0.04192895889282226
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,8,32,power_law_1.01,0.5705382537841797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,8,32,power_law_1.01,0.06209919929504395
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,8,32,power_law_1.01,0.06197120189666748
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,8,32,power_law_1.01,0.06451839923858642
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,8,32,power_law_1.01,0.06540800094604492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,8,32,power_law_1.01,0.876495361328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,8,32,power_law_1.01,0.0659609603881836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,8,32,power_law_1.01,0.06687232017517089
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,8,32,power_law_1.01,0.07104383945465088
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,8,32,power_law_1.01,1.0770022583007812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,8,32,power_law_1.01,0.07350527763366699
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,8,32,power_law_1.01,0.07912064075469971
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,8,32,power_law_1.01,0.07859200000762939
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,8,32,power_law_1.01,0.0800921630859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,8,32,power_law_1.01,0.09694592475891113
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,8,32,power_law_1.01,0.11839360237121581
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,8,32,power_law_1.01,0.15772416114807128
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,8,32,power_law_1.01,0.040779519081115725
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,8,32,power_law_1.01,0.18674688339233397
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,8,32,power_law_1.01,0.2707200050354004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,8,32,power_law_1.01,0.3895232009887695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,8,32,power_law_1.01,0.06870783805847168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,8,32,power_law_1.01,0.4213081741333008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,8,32,power_law_1.2,0.06303999900817872
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,8,32,power_law_1.01,0.07728256225585937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,8,32,power_law_1.2,0.0638592004776001
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,8,32,power_law_1.2,0.0640396785736084
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,8,32,power_law_1.01,0.6818214416503906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,8,32,power_law_1.2,0.06234879970550537
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,8,32,power_law_1.01,0.09312512397766112
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,8,32,power_law_1.2,0.10736000061035156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,8,32,power_law_1.01,0.9121459197998046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,8,32,power_law_1.2,0.10849663734436035
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,8,32,power_law_1.2,0.11350144386291502
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,8,32,power_law_1.2,0.10961664199829102
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,8,32,power_law_1.2,0.1222655963897705
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,8,32,power_law_1.2,0.12684927940368654
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,8,32,power_law_1.2,0.12747391700744629
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,8,32,power_law_1.2,0.12776703834533693
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,8,32,power_law_1.2,0.1340390396118164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,8,32,power_law_1.2,0.15024767875671388
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,8,32,power_law_1.2,0.16642816543579103
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,8,32,power_law_1.2,0.06302591800689697
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,8,32,power_law_1.2,0.22665344238281251
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,8,32,power_law_1.2,0.31627775192260743
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,8,32,power_law_1.2,0.10648832321166993
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,8,32,power_law_1.2,0.1174067211151123
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,8,32,power_law_1.2,0.12499199867248537
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,8,32,power_law_1.2,0.5998694229125976
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,8,32,power_law_1.2,0.6769856262207031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,8,32,power_law_1.2,0.040110077857971195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,8,32,power_law_1.2,0.04084864139556885
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,8,32,power_law_1.2,0.19754240036010742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,8,32,power_law_1.2,0.04154496192932129
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,8,32,power_law_1.2,0.04186367988586426
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,8,32,power_law_1.2,0.06084095954895019
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,8,32,power_law_1.2,1.080241928100586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,8,32,power_law_1.2,0.3644646453857422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,8,32,power_law_1.2,0.06718848228454591
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,8,32,power_law_1.2,0.06471168041229249
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,8,32,power_law_1.2,0.06482431888580323
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,8,32,power_law_1.2,0.0686079978942871
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,8,32,power_law_1.2,0.07229568004608154
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,8,32,power_law_1.2,0.0767961597442627
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,8,32,power_law_1.2,0.07655935764312745
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,8,32,power_law_1.2,0.07483903884887695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,8,32,power_law_1.2,0.0779750394821167
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,8,32,power_law_1.2,1.4827122497558594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,8,32,power_law_1.2,0.08123776435852051
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,8,32,power_law_1.2,0.09911423683166504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,8,32,power_law_1.2,0.11233152389526366
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,8,32,power_law_1.2,0.14089983940124512
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,8,32,power_law_1.2,0.1711359977722168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,8,32,power_law_1.2,0.24483455657958983
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,8,32,power_law_1.2,0.06258175849914552
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,8,32,power_law_1.2,0.33281150817871097
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,16,16,balanced,0.02988287925720215
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,16,16,balanced,0.029976320266723634
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,8,32,power_law_1.2,0.0659494400024414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,16,16,balanced,0.031032319068908694
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,16,16,balanced,0.03262207984924316
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,8,32,power_law_1.2,0.5397427368164063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,16,16,balanced,0.056755199432373046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,16,16,balanced,0.06738431930541992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,16,16,balanced,0.11321215629577637
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,16,16,balanced,0.11293439865112305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,16,16,balanced,0.11832703590393065
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,16,16,balanced,0.117958402633667
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,16,16,balanced,0.12036607742309571
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,16,16,balanced,0.12492799758911133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,16,16,balanced,0.12797183990478517
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,8,32,power_law_1.2,0.9106969451904297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,16,16,balanced,0.12136832237243653
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,16,16,balanced,0.12440320014953614
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,16,16,balanced,0.12775424003601074
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,16,16,balanced,0.13483519554138185
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,16,16,balanced,0.14341504096984864
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,16,16,balanced,0.17457151412963867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,16,16,balanced,0.15288576126098632
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,16,16,balanced,0.1950079917907715
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,16,16,balanced,0.2387980842590332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,8,32,power_law_1.2,1.232052459716797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,16,16,balanced,0.28018688201904296
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,16,16,balanced,0.029889280796051028
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,16,16,balanced,0.3633049774169922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,16,16,balanced,0.030082559585571288
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,16,16,balanced,0.44994430541992186
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,16,16,balanced,0.036849920749664304
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,16,16,balanced,0.04982016086578369
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,16,16,balanced,0.6257036972045898
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,16,16,balanced,0.07443583965301513
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,16,16,balanced,0.07339776039123536
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,16,16,balanced,0.0751193618774414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,16,16,balanced,0.8335782623291015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,16,16,balanced,0.07565951824188233
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,16,16,balanced,0.08147968292236328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,16,16,balanced,0.08428544044494629
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,16,16,balanced,0.08152447700500488
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,16,16,balanced,0.07750016212463379
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,16,16,balanced,0.08357248306274415
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,16,16,balanced,0.08875264167785644
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,16,16,balanced,0.09698304176330566
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,16,16,balanced,0.10722047805786132
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,16,16,balanced,0.1280076789855957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,16,16,balanced,0.1484607982635498
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,16,16,balanced,0.03048703908920288
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,16,16,balanced,0.18952064514160155
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,16,16,balanced,0.03241472005844116
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,16,16,balanced,0.3210918426513672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,16,16,power_law_1.01,0.054090237617492674
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,16,16,balanced,0.0785472011566162
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,16,16,power_law_1.01,0.06348800182342529
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,16,16,balanced,0.5699647903442383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,16,16,power_law_1.01,0.06371456146240234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,16,16,power_law_1.01,0.0644710397720337
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,16,16,balanced,0.7399565124511719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,16,16,power_law_1.01,0.10782591819763183
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,8,32,power_law_1.2,0.39746047973632814
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,16,16,power_law_1.01,0.11149184226989746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,16,16,power_law_1.01,0.1150425624847412
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,16,16,power_law_1.01,0.11673215866088867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,16,16,power_law_1.01,0.12377599716186523
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,16,16,power_law_1.01,0.1280128002166748
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,16,16,balanced,0.39615745544433595
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,16,16,power_law_1.01,0.1257420825958252
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,16,16,power_law_1.01,0.12933504104614257
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,16,16,power_law_1.01,0.13681920051574709
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,16,16,power_law_1.01,0.15114496231079103
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,16,16,power_law_1.01,0.06710912227630615
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,16,16,power_law_1.01,0.16328960418701172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,16,16,power_law_1.01,0.10857343673706055
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,16,16,power_law_1.01,0.19051647186279297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,16,16,power_law_1.01,0.23138048171997072
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,16,16,balanced,0.22806783676147463
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,16,16,power_law_1.01,0.26199296951293943
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,16,16,power_law_1.01,0.12387071609497072
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,16,16,power_law_1.01,0.35195648193359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,16,16,power_law_1.01,0.12130687713623048
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,16,16,power_law_1.01,0.03488895893096924
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,16,16,power_law_1.01,0.44156417846679685
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,16,16,power_law_1.01,0.046049280166625975
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,16,16,power_law_1.01,0.04647295951843262
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,16,16,power_law_1.01,0.04646016120910644
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,16,16,power_law_1.01,0.048947200775146485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,16,16,power_law_1.01,0.06956416130065918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,16,16,power_law_1.01,0.07213568210601806
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,16,16,power_law_1.01,0.8899878692626952
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,16,16,power_law_1.01,0.07218560218811035
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,16,16,power_law_1.01,0.07210368156433106
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,16,16,power_law_1.01,0.07543551921844482
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,16,16,power_law_1.01,0.07924608230590821
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,16,16,power_law_1.01,0.07926655769348144
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,16,16,power_law_1.01,0.0841318416595459
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,16,16,power_law_1.01,0.07958655834197999
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,16,16,power_law_1.01,1.202783966064453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,16,16,power_law_1.01,0.08445568084716797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,16,16,power_law_1.01,0.0894438362121582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,16,16,power_law_1.01,0.10852479934692383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,16,16,power_law_1.01,0.12031359672546386
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,16,16,power_law_1.01,0.18625919342041014
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,16,16,power_law_1.01,0.14632320404052734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,16,16,power_law_1.01,0.6373081588745118
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,16,16,power_law_1.01,0.2515609550476074
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,16,16,power_law_1.01,0.29390592575073243
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,16,16,power_law_1.2,0.05457536220550537
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,16,16,power_law_1.2,0.06356351852416993
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,16,16,power_law_1.01,0.4358156967163086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,16,16,power_law_1.2,0.06303103923797607
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,16,16,power_law_1.01,0.5306547164916993
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,16,16,power_law_1.2,0.06507008075714112
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,16,16,power_law_1.01,0.08202624320983887
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,16,16,power_law_1.2,0.06963200092315673
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,16,16,power_law_1.01,0.7482969665527344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,16,16,power_law_1.2,0.11103232383728028
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,16,16,power_law_1.2,0.10928895950317383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,16,16,power_law_1.2,0.11339903831481934
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,16,16,power_law_1.2,0.11741951942443847
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,16,16,power_law_1.2,0.12286335945129394
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,16,16,power_law_1.2,0.12177536010742188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,16,16,power_law_1.2,0.12865023612976073
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,16,16,power_law_1.01,1.0479065704345705
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,16,16,power_law_1.2,0.1337113571166992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,16,16,power_law_1.2,0.13600383758544923
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,16,16,power_law_1.2,0.1566489601135254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,16,16,power_law_1.2,0.17337087631225584
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,16,16,power_law_1.2,0.2055398368835449
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,16,16,power_law_1.2,0.2576460838317871
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,16,16,power_law_1.2,0.3117132759094238
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,16,16,power_law_1.2,0.37324287414550783
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,16,16,power_law_1.2,0.11357695579528808
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,16,16,power_law_1.2,0.5956825637817383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,16,16,power_law_1.2,0.03476351976394653
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,16,16,power_law_1.2,0.12774399757385252
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,16,16,power_law_1.2,0.04612736225128174
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,16,16,power_law_1.2,0.7309542083740235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,16,16,power_law_1.2,0.04455167770385742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,16,16,power_law_1.2,0.04673408031463623
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,16,16,power_law_1.2,0.04910463809967041
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,16,16,power_law_1.2,0.07074944019317628
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,16,16,power_law_1.2,0.07053184032440185
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,16,16,power_law_1.2,0.07335552215576172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,16,16,power_law_1.2,0.07809663772583007
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,16,16,power_law_1.2,0.08323712348937988
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,16,16,power_law_1.2,0.08451328277587891
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,16,16,power_law_1.2,0.12068479537963868
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,16,16,power_law_1.2,0.08000767707824707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,16,16,power_law_1.2,0.08411647796630858
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,16,16,power_law_1.2,0.0884825611114502
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,16,16,power_law_1.2,1.5662924194335939
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,16,16,power_law_1.2,0.08935296058654785
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,16,16,power_law_1.2,0.1152128028869629
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,16,16,power_law_1.2,0.12238207817077637
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,16,16,power_law_1.2,0.16297216415405275
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,16,16,power_law_1.2,0.21529983520507812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,16,16,power_law_1.2,0.07076608180999756
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,16,16,power_law_1.2,1.099015655517578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,16,16,power_law_1.2,0.3851942443847656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,32,8,balanced,0.03012864112854004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,16,16,power_law_1.2,0.5151283264160156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,32,8,balanced,0.03537152051925659
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,32,8,balanced,0.05080319881439209
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,16,16,power_law_1.2,0.6759923553466797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,32,8,balanced,0.06331264019012452
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,32,8,balanced,0.07846271991729736
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,32,8,balanced,0.12485247611999513
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,32,8,balanced,0.12217344284057616
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,32,8,balanced,0.1279155158996582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,32,8,balanced,0.13111680030822753
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,16,16,power_law_1.2,1.0335104370117187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,16,16,power_law_1.2,0.2832499122619629
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,32,8,balanced,0.1235929584503174
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,16,16,power_law_1.2,0.07391232013702392
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,32,8,balanced,0.12709376335144043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,16,16,power_law_1.2,1.1979519653320314
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,32,8,balanced,0.13555328369140623
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,32,8,balanced,0.140698881149292
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,32,8,balanced,0.1457190418243408
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,32,8,balanced,0.15359999656677245
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,32,8,balanced,0.1739366340637207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,32,8,balanced,0.19287935256958008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,32,8,balanced,0.2339878463745117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,32,8,balanced,0.2740966415405274
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,32,8,balanced,0.3579238510131836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,32,8,balanced,0.43772159576416014
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,32,8,balanced,0.030718719959259032
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,32,8,balanced,0.031065599918365477
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,32,8,balanced,0.1301632022857666
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,32,8,balanced,0.033964800834655764
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,32,8,balanced,0.03129472017288208
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,32,8,balanced,0.03824383974075317
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,32,8,balanced,1.0982412719726562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,32,8,balanced,0.04540671825408936
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,32,8,balanced,0.06581247806549072
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,32,8,balanced,0.09422207832336425
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,32,8,balanced,0.09912960052490234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,32,8,balanced,1.4613183593749999
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,32,8,balanced,0.09650815963745117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,32,8,balanced,0.6005286407470704
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,32,8,balanced,0.09874176025390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,32,8,balanced,0.10376447677612304
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,32,8,balanced,0.76640380859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,32,8,balanced,0.10683135986328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,32,8,balanced,0.11179264068603516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,32,8,balanced,0.1223680019378662
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,32,8,balanced,0.14191231727600098
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,32,8,balanced,0.1632934379577637
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,32,8,balanced,0.20227455139160155
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,32,8,balanced,0.09300095558166503
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,32,8,balanced,0.24038528442382812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,32,8,balanced,0.3194867134094238
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,32,8,balanced,0.39346942901611326
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,32,8,balanced,0.09269248008728027
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,32,8,power_law_1.01,0.033766400814056394
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,32,8,balanced,0.5781427383422851
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,32,8,power_law_1.01,0.05736703872680664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,32,8,power_law_1.01,0.06350592136383057
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,32,8,balanced,0.7306355285644531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,32,8,power_law_1.01,0.07231488227844238
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,32,8,power_law_1.01,0.07478528022766114
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,32,8,power_law_1.01,0.11580287933349609
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,32,8,power_law_1.01,0.11964799880981444
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,32,8,balanced,1.0694080352783204
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,32,8,power_law_1.01,0.12448127746582031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,32,8,power_law_1.01,0.12928640365600585
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,32,8,balanced,0.09684864044189453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,32,8,power_law_1.01,0.12929535865783692
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,32,8,power_law_1.01,0.12581631660461426
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,32,8,balanced,1.4097984313964844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,32,8,power_law_1.01,0.13171968460083008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,32,8,power_law_1.01,0.1361881637573242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,32,8,power_law_1.01,0.14312447547912596
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,32,8,power_law_1.01,0.16438911437988282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,32,8,power_law_1.01,0.1867033576965332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,32,8,power_law_1.01,0.22083200454711913
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,32,8,power_law_1.01,0.2518617630004883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,32,8,power_law_1.01,0.3142412757873535
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,32,8,power_law_1.01,0.12293120384216309
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,32,8,power_law_1.01,0.3968703842163086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,32,8,power_law_1.01,0.5252710342407226
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,32,8,power_law_1.01,0.03535360097885132
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,32,8,power_law_1.01,0.1282316780090332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,32,8,power_law_1.01,0.7103692626953124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,32,8,power_law_1.01,0.05204864025115967
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,32,8,power_law_1.01,0.04929152011871338
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,32,8,power_law_1.01,0.0555072021484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,32,8,power_law_1.01,0.06521471977233886
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,32,8,power_law_1.01,0.061631999015808105
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,32,8,power_law_1.01,1.0278988647460938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,32,8,power_law_1.01,0.08728832244873047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,32,8,power_law_1.01,0.08949631690979004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,32,8,power_law_1.01,0.09158783912658691
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,32,8,power_law_1.01,0.09464575767517089
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,32,8,power_law_1.01,0.09774080276489258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,32,8,power_law_1.01,1.3081805419921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,32,8,power_law_1.01,0.09499263763427734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,32,8,power_law_1.01,0.09761024475097656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,32,8,power_law_1.01,0.10076288223266601
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,32,8,power_law_1.01,0.10995072364807128
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,32,8,power_law_1.01,0.04178304195404053
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,32,8,power_law_1.01,0.15543680191040038
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,32,8,power_law_1.01,0.13498623847961427
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,32,8,power_law_1.01,0.18409984588623046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,32,8,power_law_1.01,0.21465087890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,32,8,power_law_1.01,0.3161356735229492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,32,8,power_law_1.01,0.0866534423828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,32,8,power_law_1.01,0.3780518341064453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,32,8,power_law_1.2,0.03405312061309814
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,32,8,power_law_1.01,0.5072217559814453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,32,8,power_law_1.01,0.09201151847839355
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,32,8,power_law_1.01,0.7070579528808594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,32,8,power_law_1.2,0.059966721534729005
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,32,8,power_law_1.2,0.06361216068267822
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,32,8,power_law_1.2,0.06985087871551514
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,32,8,power_law_1.2,0.11955967903137207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,32,8,power_law_1.01,0.9633663940429689
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,32,8,power_law_1.2,0.11858176231384276
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,32,8,power_law_1.2,0.12243455886840819
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,32,8,power_law_1.2,0.12507776260375975
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,32,8,power_law_1.2,0.12661120414733887
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,32,8,power_law_1.2,0.12720255851745604
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,32,8,power_law_1.2,0.12792832374572755
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,32,8,power_law_1.2,0.13373056411743164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,32,8,power_law_1.2,0.13676544189453127
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,32,8,power_law_1.2,0.1469491195678711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,32,8,power_law_1.2,0.16248960494995118
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,32,8,power_law_1.2,0.1814860725402832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,32,8,power_law_1.2,0.07703680038452149
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,32,8,power_law_1.2,0.2350720024108887
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,32,8,power_law_1.2,0.2654579162597656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,32,8,power_law_1.2,0.12931072235107421
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,32,8,power_law_1.01,1.3329689025878906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,32,8,power_law_1.2,0.4562060928344726
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,32,8,power_law_1.2,0.058307838439941403
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,32,8,power_law_1.2,0.03538687944412232
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,32,8,power_law_1.2,0.041985278129577634
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,32,8,power_law_1.2,0.7812992095947265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,32,8,power_law_1.2,0.04508927822113037
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,32,8,power_law_1.2,0.05092095851898193
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,32,8,power_law_1.2,0.05697279930114746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,32,8,power_law_1.2,0.34765441894531246
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,32,8,power_law_1.2,0.08630911827087402
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,32,8,power_law_1.2,0.08660351753234863
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,32,8,power_law_1.2,1.2664025878906249
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,32,8,power_law_1.2,0.0891481590270996
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,32,8,power_law_1.2,0.09072768211364746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,32,8,power_law_1.2,0.5574604797363281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,32,8,power_law_1.2,1.4399270629882812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,32,8,power_law_1.2,0.09609215736389161
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,32,8,power_law_1.2,0.09829631805419922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,32,8,power_law_1.2,0.09467391967773438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,32,8,power_law_1.2,0.09554047584533691
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,32,8,power_law_1.2,0.09896575927734375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,32,8,power_law_1.2,0.10147711753845215
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,32,8,power_law_1.2,0.13940223693847656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,32,8,power_law_1.2,0.1589951992034912
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,32,8,power_law_1.2,0.19854080200195312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,32,8,power_law_1.2,0.24348800659179687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,32,8,power_law_1.2,0.3156889533996582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,32,8,power_law_1.2,0.06358399868011474
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,32,8,power_law_1.2,0.39143680572509765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,1,256,balanced,0.06125440120697021
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,1,256,balanced,0.06176511764526367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,1,256,balanced,0.06170368194580078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,32,8,power_law_1.2,0.5595929718017578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,1,256,balanced,0.06129536151885986
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,1,256,balanced,0.061452798843383796
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,1,256,balanced,0.06209536075592041
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,1,256,balanced,0.1107750415802002
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,1,256,balanced,0.11164031982421876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,32,8,power_law_1.2,0.7705010986328125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,1,256,balanced,0.11100543975830077
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,1,256,balanced,0.11104512214660645
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,1,256,balanced,0.11066368103027344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,32,8,power_law_1.2,0.11271679878234864
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,1,256,balanced,0.11163007736206054
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,1,256,balanced,0.11333888053894044
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,32,8,power_law_1.2,1.109365768432617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,1,256,balanced,0.11444992065429686
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,1,256,balanced,0.11557632446289064
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,1,256,balanced,0.12137855529785155
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,1,256,balanced,0.12848511695861817
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,1,256,balanced,0.13398271560668945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,1,256,balanced,0.13217535972595215
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,1,256,balanced,0.12886272430419923
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,1,256,balanced,0.13844991683959962
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,1,256,balanced,0.14863360404968262
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,1,256,balanced,0.18211200714111328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,1,256,balanced,0.03350528001785279
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,1,256,balanced,0.21859455108642578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,1,256,balanced,0.032385280132293706
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,1,256,balanced,0.033575680255889896
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,1,256,balanced,0.03237375974655151
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,1,256,balanced,0.111724796295166
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,1,256,balanced,0.11289216041564942
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,1,256,balanced,0.05537536144256592
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,1,256,balanced,0.05520256042480469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,1,256,balanced,0.11347071647644043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,1,256,balanced,0.05662335872650146
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,1,256,balanced,0.055823359489440914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,1,256,balanced,0.055694079399108885
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,1,256,balanced,0.05613056182861328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,1,256,balanced,0.05650047779083252
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,1,256,balanced,0.05746816158294678
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,1,256,balanced,0.0578329610824585
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,1,256,balanced,0.05760511875152587
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,1,256,balanced,0.061594882011413575
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,1,256,balanced,0.06553343772888184
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,1,256,balanced,0.06497151851654052
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,1,256,balanced,0.06841728210449219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,1,256,balanced,0.032334079742431646
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,1,256,balanced,0.07461120128631592
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,1,256,balanced,0.07047935962677002
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,1,256,balanced,0.07864192008972168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,1,256,balanced,0.08592384338378907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,1,256,balanced,0.09869440078735352
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,1,256,balanced,0.03334784030914307
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,1,256,balanced,0.11511167526245117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,1,256,power_law_1.01,0.061761279106140134
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,1,256,power_law_1.01,0.061378560066223144
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,1,256,power_law_1.01,0.06163968086242676
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,1,256,power_law_1.01,0.11171584129333496
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,1,256,balanced,0.05761536121368408
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,32,8,power_law_1.2,1.3861772155761718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,1,256,power_law_1.01,0.11087615966796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,1,256,power_law_1.01,0.11145343780517578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,1,256,power_law_1.01,0.11397248268127443
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,1,256,power_law_1.01,0.11579648017883301
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,1,256,power_law_1.01,0.11609087944030763
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,1,256,power_law_1.01,0.12166399955749511
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,1,256,power_law_1.01,0.11926527976989747
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,1,256,power_law_1.01,0.12317567825317384
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,1,256,power_law_1.01,0.12583040237426757
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,1,256,power_law_1.01,0.12586496353149415
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,1,256,power_law_1.01,0.14178815841674805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,1,256,power_law_1.01,0.17400192260742187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,1,256,power_law_1.01,0.06222847938537598
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,1,256,power_law_1.01,0.21675647735595702
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,1,256,power_law_1.01,0.3225126266479492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,1,256,power_law_1.01,0.37528190612792967
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,1,256,power_law_1.01,0.5432511901855468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,1,256,power_law_1.01,0.1313100814819336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,1,256,power_law_1.01,0.032767999172210696
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,1,256,power_law_1.01,0.8219929504394532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,1,256,power_law_1.01,0.03280639886856079
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,1,256,power_law_1.01,0.03247488021850586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,1,256,power_law_1.01,0.03363840103149414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,1,256,power_law_1.01,0.05532288074493408
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,1,256,power_law_1.01,0.055731201171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,1,256,power_law_1.01,0.056112642288208
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,1,256,power_law_1.01,0.05719935894012451
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,1,256,power_law_1.01,1.320885772705078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,1,256,power_law_1.01,0.058280959129333496
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,1,256,power_law_1.01,0.058406400680541995
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,1,256,power_law_1.01,0.06164735794067383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,1,256,power_law_1.01,0.06458879947662353
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,1,256,power_law_1.01,1.6325299072265627
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,1,256,power_law_1.01,0.13128959655761718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,1,256,power_law_1.01,0.06771071910858154
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,1,256,power_law_1.01,0.07667456150054933
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,1,256,power_law_1.01,0.07484416007995606
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,1,256,power_law_1.01,0.08046208381652832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,1,256,power_law_1.01,0.11196672439575195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,1,256,power_law_1.01,0.11051903724670412
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,1,256,power_law_1.01,0.1429811191558838
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,1,256,power_law_1.01,0.1912563133239746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,1,256,power_law_1.01,0.058048000335693364
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,1,256,power_law_1.01,0.256942081451416
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,1,256,power_law_1.2,0.06169983863830566
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,1,256,power_law_1.01,0.3599795150756836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,1,256,power_law_1.2,0.06164864063262939
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,1,256,power_law_1.01,0.06373504161834717
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,1,256,power_law_1.01,0.06663680076599121
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,1,256,power_law_1.2,0.061815037727355956
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,1,256,power_law_1.01,0.580972785949707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,1,256,power_law_1.2,0.1118182373046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,1,256,power_law_1.2,0.11145088195800781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,1,256,power_law_1.2,0.11144448280334474
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,1,256,power_law_1.2,0.11149951934814453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,1,256,power_law_1.2,0.11357952117919923
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,1,256,power_law_1.2,0.11639039993286134
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,1,256,power_law_1.2,0.11509632110595702
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,1,256,power_law_1.2,0.11607423782348633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,1,256,power_law_1.2,0.12067199707031251
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,1,256,power_law_1.2,0.12055295944213866
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,1,256,power_law_1.2,0.12596991539001465
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,1,256,power_law_1.2,0.12901887893676758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,1,256,power_law_1.2,0.13210495948791504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,1,256,power_law_1.2,0.13470208168029785
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,1,256,power_law_1.2,0.15014783859252928
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,1,256,power_law_1.2,0.17718271255493162
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,1,256,power_law_1.2,0.2457356834411621
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,1,256,power_law_1.2,0.3171622467041016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,1,256,power_law_1.2,0.4913318252563476
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,1,256,power_law_1.01,0.7908415985107422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,1,256,power_law_1.2,0.6680973052978516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,1,256,power_law_1.2,0.03250688076019287
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,1,256,power_law_1.2,0.032446720600128175
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,1,256,power_law_1.2,0.032616961002349856
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,1,256,power_law_1.2,0.05569920063018798
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,1,256,power_law_1.2,1.1583744049072267
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,1,256,power_law_1.2,0.05576704025268555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,1,256,power_law_1.2,0.05622655868530273
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,1,256,power_law_1.2,0.05727871894836426
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,1,256,power_law_1.2,1.4720614624023436
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,1,256,power_law_1.2,0.05661183834075928
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,1,256,power_law_1.2,0.058283519744873044
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,1,256,power_law_1.2,0.05873023986816406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,1,256,power_law_1.2,0.06291840076446534
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,1,256,power_law_1.01,0.09195008277893066
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,1,256,power_law_1.2,0.06332799911499024
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,1,256,power_law_1.2,0.06344064235687255
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,1,256,power_law_1.2,0.06631936073303223
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,1,256,power_law_1.2,0.07292160034179687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,1,256,power_law_1.2,0.07718656063079835
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,1,256,power_law_1.2,0.07627903938293457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,1,256,power_law_1.2,0.08337151527404785
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,1,256,power_law_1.2,0.09239168167114258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,1,256,power_law_1.2,0.11894016265869141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,1,256,power_law_1.2,0.1621708869934082
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,1,256,power_law_1.2,0.23140991210937498
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,1,256,power_law_1.2,0.3108902359008789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,1,256,power_law_1.2,0.47253761291503904
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,1,256,power_law_1.2,2.821825256347656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,2,128,balanced,0.03521536111831665
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,2,128,balanced,0.044408321380615234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,2,128,balanced,0.042342400550842284
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,1,256,power_law_1.2,0.7879615783691406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,2,128,balanced,0.03001600027084351
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,1,256,power_law_1.2,3.2287783813476567
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,2,128,balanced,0.056158719062805174
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,2,128,balanced,0.09881855964660644
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,2,128,balanced,0.0994547176361084
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,2,128,balanced,0.10055808067321777
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,2,128,balanced,0.10016639709472656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,2,128,balanced,0.10248064041137694
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,2,128,balanced,0.10387583732604981
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,2,128,balanced,0.10333567619323732
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,2,128,balanced,0.038309121131896974
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,2,128,balanced,0.10570240020751953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,2,128,balanced,0.10967424392700195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,2,128,balanced,0.10928640365600586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,2,128,balanced,0.11519743919372558
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,2,128,balanced,0.12139007568359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,2,128,balanced,0.11499263763427733
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,1,256,power_law_1.2,1.793190460205078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,2,128,balanced,0.12292991638183595
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,2,128,balanced,0.1300543975830078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,2,128,balanced,0.14418944358825683
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,2,128,balanced,0.09902463912963867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,2,128,balanced,0.15647487640380858
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,2,128,balanced,0.23770879745483398
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,2,128,balanced,0.02935424089431763
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,2,128,balanced,0.029587199687957765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,1,256,power_law_1.2,1.1944281768798828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,2,128,balanced,0.029437439441680906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,2,128,balanced,0.02938112020492554
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,2,128,balanced,0.0513420820236206
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,2,128,balanced,0.05071872234344482
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,2,128,balanced,0.050749440193176266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,2,128,balanced,0.05261824131011963
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,2,128,balanced,0.05177343845367431
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,2,128,balanced,0.051893758773803714
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,2,128,balanced,0.10627967834472656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,2,128,balanced,0.052391681671142575
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,2,128,balanced,0.0578111982345581
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,2,128,balanced,0.056524801254272464
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,2,128,balanced,0.05820543766021728
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,2,128,balanced,0.06424448013305664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,2,128,balanced,0.07007999897003174
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,2,128,balanced,0.06300159931182861
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,2,128,balanced,0.07045631885528565
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,2,128,balanced,0.19643264770507812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,2,128,balanced,0.07624447822570801
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,2,128,balanced,0.08800512313842773
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,2,128,balanced,0.1001471996307373
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,2,128,balanced,0.12455807685852052
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,2,128,balanced,0.029945600032806396
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,2,128,balanced,0.14939135551452637
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,2,128,power_law_1.01,0.05678592205047608
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,2,128,power_law_1.01,0.05696000099182129
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,2,128,power_law_1.01,0.05621503829956055
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,2,128,power_law_1.01,0.09775487899780273
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,2,128,power_law_1.01,0.10134400367736815
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,2,128,balanced,0.05630080223083496
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,2,128,balanced,0.057715201377868654
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,2,128,power_law_1.01,0.10102272033691406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,2,128,power_law_1.01,0.10401535987854005
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,2,128,power_law_1.01,0.10808447837829589
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,2,128,power_law_1.01,0.10444160461425782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,2,128,power_law_1.01,0.10635519981384278
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,2,128,balanced,0.02924799919128418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,2,128,power_law_1.01,0.1152780818939209
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,2,128,power_law_1.01,0.11899904251098632
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,2,128,power_law_1.01,0.12302207946777344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,2,128,power_law_1.01,0.12531200408935547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,2,128,power_law_1.01,0.056332798004150385
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,2,128,power_law_1.01,0.13553279876708985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,2,128,power_law_1.01,0.14875519752502442
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,2,128,power_law_1.01,0.1735807991027832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,2,128,power_law_1.01,0.2567334365844727
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,2,128,power_law_1.01,0.35272449493408203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,2,128,power_law_1.01,0.10291839599609376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,2,128,power_law_1.01,0.49602432250976564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,2,128,power_law_1.01,0.03010175943374634
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,2,128,power_law_1.01,0.10750975608825683
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,2,128,power_law_1.01,0.030394880771636962
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,2,128,power_law_1.01,0.6983347320556641
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,2,128,power_law_1.01,0.030088961124420166
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,2,128,power_law_1.01,0.030210559368133542
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,2,128,power_law_1.01,0.11841919898986816
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,2,128,power_law_1.01,0.0508134412765503
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,2,128,power_law_1.01,0.052773118019104004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,2,128,power_law_1.01,0.9825881958007813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,2,128,power_law_1.01,0.05281792163848877
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,2,128,power_law_1.01,0.05354623794555664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,2,128,power_law_1.01,0.05332863807678223
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,2,128,power_law_1.01,0.060522241592407225
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,2,128,power_law_1.01,0.05908607959747314
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,2,128,power_law_1.01,0.06216447830200196
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,2,128,power_law_1.01,0.06361472129821777
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,2,128,power_law_1.01,0.06634111881256104
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,2,128,power_law_1.01,0.07015935897827148
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,2,128,power_law_1.01,1.4957836914062501
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,2,128,power_law_1.01,0.0720576000213623
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,2,128,power_law_1.01,0.07761023998260498
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,2,128,power_law_1.01,0.0949452781677246
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,2,128,power_law_1.01,0.09980928421020507
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,2,128,power_law_1.01,0.15622655868530272
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,2,128,power_law_1.01,0.1371020793914795
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,2,128,power_law_1.01,0.28225664138793943
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,2,128,power_law_1.01,0.05214975833892822
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,2,128,power_law_1.2,0.056206078529357904
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,2,128,power_law_1.01,0.057715201377868654
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,2,128,power_law_1.2,0.05629183769226074
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,2,128,power_law_1.01,0.5241907119750977
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,2,128,power_law_1.2,0.0982425594329834
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,2,128,power_law_1.2,0.09737983703613282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,2,128,power_law_1.2,0.10169343948364258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,2,128,power_law_1.01,0.7829183959960938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,2,128,power_law_1.2,0.1012992000579834
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,2,128,power_law_1.2,0.10408063888549805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,2,128,power_law_1.2,0.10637056350708007
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,2,128,power_law_1.2,0.10912256240844727
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,2,128,power_law_1.2,0.10595328330993652
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,2,128,power_law_1.2,0.11191679954528808
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,2,128,power_law_1.2,0.1159385585784912
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,2,128,power_law_1.01,0.34103038787841794
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,2,128,power_law_1.2,0.12372608184814453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,2,128,power_law_1.2,0.05643648147583008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,2,128,power_law_1.2,0.12231040000915527
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,2,128,power_law_1.2,0.1357913589477539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,2,128,power_law_1.2,0.14967679977416992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,2,128,power_law_1.2,0.1852672004699707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,2,128,power_law_1.2,0.25247232437133793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,2,128,power_law_1.2,0.1026585578918457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,2,128,power_law_1.2,0.3902988815307617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,2,128,power_law_1.2,0.4617932891845703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,2,128,power_law_1.2,0.03004928112030029
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,2,128,power_law_1.2,0.029911038875579832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,2,128,power_law_1.2,0.03169536113739014
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,2,128,power_law_1.2,0.11920512199401856
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,2,128,power_law_1.2,0.05059840202331543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,2,128,power_law_1.2,0.05071487903594971
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,2,128,power_law_1.2,0.051765761375427245
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,2,128,power_law_1.2,1.0886886596679688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,2,128,power_law_1.2,0.053350400924682614
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,2,128,power_law_1.2,0.05271423816680908
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,2,128,power_law_1.2,0.05513343811035156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,2,128,power_law_1.2,0.06052095890045166
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,2,128,power_law_1.2,0.05950463771820068
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,2,128,power_law_1.2,0.06390783786773682
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,2,128,power_law_1.2,1.548528594970703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,2,128,power_law_1.2,0.06975103855133057
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,2,128,power_law_1.2,0.06875648021697997
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,2,128,power_law_1.2,0.06707968235015868
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,2,128,power_law_1.2,0.0872537612915039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,2,128,power_law_1.2,0.8292134094238282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,2,128,power_law_1.2,0.1104319953918457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,2,128,power_law_1.2,0.1337996768951416
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,2,128,power_law_1.2,0.18270336151123048
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,2,128,power_law_1.2,0.05278719902038574
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,2,128,power_law_1.2,0.2686963272094727
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,2,128,power_law_1.2,2.493928985595703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,2,128,power_law_1.2,0.05774464130401611
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,2,128,power_law_1.2,0.4275571060180664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,4,64,balanced,0.028437759876251224
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,4,64,balanced,0.02792191982269287
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,4,64,balanced,0.027783679962158202
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,4,64,balanced,0.047944960594177244
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,2,128,power_law_1.2,0.6401331329345703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,4,64,balanced,0.05552256107330322
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,4,64,balanced,0.09426943778991699
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,2,128,power_law_1.2,0.07919871807098389
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,4,64,balanced,0.09565823554992677
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,4,64,balanced,0.09503744125366212
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,4,64,balanced,0.09714431762695312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,2,128,power_law_1.2,0.9278067016601563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,4,64,balanced,0.09751935958862304
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,4,64,balanced,0.0994547176361084
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,4,64,balanced,0.10320768356323243
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,4,64,balanced,0.10561280250549317
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,4,64,balanced,0.10876288414001464
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,4,64,balanced,0.11383040428161621
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,4,64,balanced,0.10904064178466796
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,4,64,balanced,0.11572223663330079
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,2,128,power_law_1.2,1.3768435668945311
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,4,64,balanced,0.1218393611907959
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,4,64,balanced,0.1341055965423584
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,4,64,balanced,0.14397055625915528
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,4,64,balanced,0.02835968017578125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,4,64,balanced,0.1684147262573242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,4,64,balanced,0.18986495971679687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,4,64,balanced,0.24395008087158204
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,4,64,balanced,0.028056321144104
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,4,64,balanced,0.30680831909179684
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,4,64,balanced,0.02831615924835205
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,4,64,balanced,0.02817408084869385
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,4,64,balanced,0.028879361152648924
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,4,64,balanced,0.0306329607963562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,4,64,balanced,0.05089663982391357
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,4,64,balanced,0.09729023933410644
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,4,64,balanced,0.05089663982391357
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,4,64,balanced,0.05355519771575927
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,4,64,balanced,0.052378878593444825
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,4,64,balanced,0.05110527992248535
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,4,64,balanced,0.054451198577880854
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,4,64,balanced,0.0542848014831543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,4,64,balanced,0.06224768161773682
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,4,64,balanced,0.05756671905517578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,4,64,balanced,0.05911935806274414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,4,64,balanced,0.06115583896636963
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,4,64,balanced,0.06803071975708008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,4,64,balanced,0.06135680198669433
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,4,64,balanced,0.06855167865753173
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,4,64,balanced,0.07308288097381592
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,4,64,balanced,0.08102399826049804
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,4,64,balanced,0.09185024261474609
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,4,64,balanced,0.11707519531249999
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,4,64,balanced,0.13620351791381835
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,4,64,balanced,0.028330240249633785
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,4,64,balanced,0.1776473617553711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,4,64,balanced,0.09518336296081543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,4,64,balanced,0.2202662467956543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,4,64,power_law_1.01,0.05498623847961426
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,4,64,power_law_1.01,0.055592961311340325
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,4,64,power_law_1.01,0.055852799415588385
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,4,64,power_law_1.01,0.05557631969451904
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,4,64,power_law_1.01,0.09285120010375977
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,4,64,power_law_1.01,0.0947929573059082
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,4,64,power_law_1.01,0.09548159599304198
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,4,64,power_law_1.01,0.09673855781555177
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,4,64,power_law_1.01,0.10031488418579101
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,4,64,power_law_1.01,0.09740544319152832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,4,64,power_law_1.01,0.10185215950012208
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,4,64,power_law_1.01,0.1059455966949463
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,4,64,power_law_1.01,0.10828927993774413
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,4,64,power_law_1.01,0.11393152236938478
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,4,64,power_law_1.01,0.11445504188537599
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,4,64,power_law_1.01,0.11526528358459473
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,4,64,power_law_1.01,0.11671168327331542
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,4,64,power_law_1.01,0.12470911979675292
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,4,64,power_law_1.01,0.13346943855285645
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,4,64,power_law_1.01,0.148787202835083
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,4,64,power_law_1.01,0.1639423942565918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,4,64,power_law_1.01,0.22623615264892577
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,4,64,power_law_1.01,0.2902387237548828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,4,64,power_law_1.01,0.030293760299682615
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,4,64,power_law_1.01,0.4168691253662109
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,4,64,power_law_1.01,0.03039103984832764
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,4,64,power_law_1.01,0.03232640027999878
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,4,64,power_law_1.01,0.030686719417572023
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,4,64,power_law_1.01,0.051490559577941894
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,4,64,power_law_1.01,0.05000192165374756
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,4,64,power_law_1.01,0.758787841796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,4,64,power_law_1.01,0.056179199218749996
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,4,64,power_law_1.01,0.0544652795791626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,4,64,power_law_1.01,0.05611392021179199
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,4,64,power_law_1.01,0.9638451385498048
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,4,64,power_law_1.01,0.061377282142639156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,4,64,power_law_1.01,0.06559360027313232
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,4,64,power_law_1.01,0.0679859209060669
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,4,64,power_law_1.01,0.06921984195709228
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,4,64,power_law_1.01,0.07476863861083985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,4,64,power_law_1.01,0.0810752010345459
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,4,64,power_law_1.01,0.10100095748901368
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,4,64,power_law_1.01,0.555601921081543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,4,64,power_law_1.01,0.11581567764282226
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,4,64,power_law_1.01,0.052334079742431636
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,4,64,power_law_1.01,0.15070464134216308
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,4,64,power_law_1.01,0.20343040466308593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,4,64,power_law_1.01,0.26421888351440426
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,4,64,power_law_1.01,0.058085122108459473
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,4,64,power_law_1.01,0.0595686388015747
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,4,64,power_law_1.01,0.06583680152893066
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,4,64,power_law_1.2,0.05544447898864746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,4,64,power_law_1.01,0.3964531326293945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,4,64,power_law_1.2,0.05524991989135743
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,4,64,power_law_1.2,0.05341184139251709
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,4,64,power_law_1.01,0.5204288101196288
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,4,64,power_law_1.2,0.0920358371734619
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,4,64,power_law_1.2,0.09518207550048828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,4,64,power_law_1.2,0.09653375625610353
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,4,64,power_law_1.2,0.09643263816833496
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,4,64,power_law_1.2,0.09772543907165528
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,4,64,power_law_1.01,0.7502642822265625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,4,64,power_law_1.2,0.10224639892578125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,4,64,power_law_1.2,0.09790335655212402
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,4,64,power_law_1.2,0.10708607673645018
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,4,64,power_law_1.2,0.1092300796508789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,4,64,power_law_1.2,0.11004544258117677
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,4,64,power_law_1.2,0.12003968238830567
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,4,64,power_law_1.2,0.11394816398620607
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,4,64,power_law_1.2,0.13299327850341797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,4,64,power_law_1.2,0.1456460762023926
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,4,64,power_law_1.2,0.18490367889404297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,4,64,power_law_1.2,0.22589311599731446
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,4,64,power_law_1.2,0.30014591217041015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,4,64,power_law_1.2,0.42069377899169924
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,4,64,power_law_1.2,0.5934195327758789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,4,64,power_law_1.2,0.030519039630889894
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,4,64,power_law_1.2,0.11494655609130859
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,4,64,power_law_1.2,0.785448989868164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,4,64,power_law_1.2,0.031741440296173096
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,4,64,power_law_1.2,0.049708800315856935
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,4,64,power_law_1.2,0.05085440158843994
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,4,64,power_law_1.2,0.052110080718994145
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,4,64,power_law_1.2,0.055418882369995114
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,4,64,power_law_1.2,1.1835123443603517
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,4,64,power_law_1.2,0.0522982406616211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,4,64,power_law_1.2,0.053454079627990723
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,4,64,power_law_1.2,0.05609087944030762
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,4,64,power_law_1.2,0.05749120235443115
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,4,64,power_law_1.2,0.05934847831726074
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,4,64,power_law_1.2,0.05849984169006348
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,4,64,power_law_1.2,1.5761447143554688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,4,64,power_law_1.2,0.06055424213409424
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,4,64,power_law_1.2,0.07013376235961914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,4,64,power_law_1.2,0.06681471824645996
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,4,64,power_law_1.2,0.06967040061950683
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,4,64,power_law_1.2,0.03045248031616211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,4,64,power_law_1.2,0.09029248237609863
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,4,64,power_law_1.2,0.08068608283996583
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,4,64,power_law_1.2,0.12328831672668457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,4,64,power_law_1.2,0.14516736030578614
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,4,64,power_law_1.2,0.20279296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,4,64,power_law_1.2,0.25893760681152345
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,8,32,balanced,0.02703999996185303
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,8,32,balanced,0.02738944053649902
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,8,32,balanced,0.027080960273742676
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,8,32,balanced,0.04675327777862549
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,8,32,balanced,0.05624703884124756
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,4,64,power_law_1.2,0.5638694381713867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,8,32,balanced,0.09436032295227051
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,4,64,power_law_1.2,0.06982272148132325
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,8,32,balanced,0.0951807975769043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,8,32,balanced,0.10068351745605468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,8,32,balanced,0.09659135818481446
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,8,32,balanced,0.10018176078796386
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,8,32,balanced,0.10330368041992188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,8,32,balanced,0.10626303672790527
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,4,64,power_law_1.2,1.0189990234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,8,32,balanced,0.11391743659973144
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,8,32,balanced,0.1061734390258789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,8,32,balanced,0.11293055534362792
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,8,32,balanced,0.11863039970397948
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,8,32,balanced,0.12912384033203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,8,32,balanced,0.02913408041000366
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,8,32,balanced,0.13831040382385254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,4,64,power_law_1.2,0.436943359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,8,32,balanced,0.15982336044311524
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,8,32,balanced,0.17782655715942383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,8,32,balanced,0.21800960540771483
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,8,32,balanced,0.2576038360595703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,4,64,power_law_1.2,0.7891340637207032
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,8,32,balanced,0.02765183925628662
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,8,32,balanced,0.343422737121582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,8,32,balanced,0.10136575698852539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,8,32,balanced,0.027731199264526368
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,8,32,balanced,0.4459276962280273
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,8,32,balanced,0.02965759992599487
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,8,32,balanced,0.028334081172943115
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,8,32,balanced,0.037363200187683104
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,8,32,balanced,0.11081727981567382
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,8,32,balanced,0.05643136024475097
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,8,32,balanced,0.05662591934204102
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,8,32,balanced,0.06264832019805908
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,8,32,balanced,0.05878015995025635
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,8,32,balanced,0.05998335838317871
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,8,32,balanced,0.05977471828460693
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,8,32,balanced,0.06131072044372558
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,8,32,balanced,0.06513792037963867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,8,32,balanced,0.067609601020813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,8,32,balanced,0.07072127819061279
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,8,32,balanced,0.06299903869628906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,8,32,balanced,0.06951680183410644
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,8,32,balanced,0.0741209602355957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,8,32,balanced,0.08149760246276856
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,8,32,balanced,0.09042816162109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,8,32,balanced,0.11134336471557618
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,8,32,balanced,0.1301478385925293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,8,32,balanced,0.17341951370239256
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,8,32,balanced,0.20884735107421876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,8,32,power_law_1.01,0.05500800132751464
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,8,32,balanced,0.2879001617431641
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,8,32,balanced,0.027944960594177247
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,8,32,power_law_1.01,0.05655935764312744
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,8,32,balanced,0.36614784240722653
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,8,32,power_law_1.01,0.05684607982635498
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,8,32,power_law_1.01,0.09277567863464356
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,8,32,power_law_1.01,0.09309823989868164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,8,32,power_law_1.01,0.09419903755187989
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,8,32,power_law_1.01,0.09536128044128418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,8,32,power_law_1.01,0.10001791954040526
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,8,32,power_law_1.01,0.09663999557495118
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,8,32,power_law_1.01,0.10792575836181642
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,8,32,power_law_1.01,0.10840191841125488
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,8,32,power_law_1.01,0.11377920150756834
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,8,32,power_law_1.01,0.10917632102966308
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,8,32,power_law_1.01,0.11198335647583009
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,8,32,power_law_1.01,0.11716608047485351
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,8,32,power_law_1.01,0.1265446376800537
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,8,32,power_law_1.01,0.1348697566986084
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,8,32,power_law_1.01,0.16356735229492186
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,8,32,power_law_1.01,0.1860518455505371
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,8,32,power_law_1.01,0.2294118309020996
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,8,32,power_law_1.01,0.055509757995605466
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,8,32,power_law_1.01,0.24942720413208005
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,8,32,power_law_1.01,0.37720703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,8,32,power_law_1.01,0.03578495979309082
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,8,32,power_law_1.01,0.44566398620605463
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,8,32,power_law_1.01,0.036599040031433105
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,8,32,power_law_1.01,0.10244607925415039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,8,32,power_law_1.01,0.03642751932144165
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,8,32,power_law_1.01,0.03730815887451172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,8,32,power_law_1.01,0.054888958930969234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,8,32,power_law_1.01,0.7399321746826172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,8,32,power_law_1.01,0.05756800174713135
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,8,32,power_law_1.01,0.05886847972869873
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,8,32,power_law_1.01,0.06160384178161621
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,8,32,power_law_1.01,0.9806963348388672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,8,32,power_law_1.01,0.0653388786315918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,8,32,power_law_1.01,0.06732672214508056
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,8,32,power_law_1.01,0.07293056011199951
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,8,32,power_law_1.01,0.06921472072601317
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,8,32,power_law_1.01,0.07164031982421874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,8,32,power_law_1.01,0.0813593578338623
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,8,32,power_law_1.01,0.08865792274475097
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,8,32,power_law_1.01,0.12486911773681642
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,8,32,power_law_1.01,0.05477759838104248
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,8,32,power_law_1.01,0.14054271697998047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,8,32,power_law_1.01,0.17797119140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,8,32,power_law_1.01,0.05825535774230957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,8,32,power_law_1.01,0.22333183288574218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,8,32,power_law_1.2,0.05507455825805664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,8,32,power_law_1.01,0.3062835121154785
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,8,32,power_law_1.01,0.39788158416748043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,8,32,power_law_1.01,0.06767104148864747
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,8,32,power_law_1.2,0.05630720138549805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,8,32,power_law_1.01,0.5773465728759766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,8,32,power_law_1.2,0.09147135734558105
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,8,32,power_law_1.2,0.09193599700927735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,8,32,power_law_1.2,0.09447680473327637
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,8,32,power_law_1.01,0.8068134307861328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,8,32,power_law_1.2,0.09980416297912598
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,8,32,power_law_1.2,0.09791359901428223
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,8,32,power_law_1.2,0.10049280166625976
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,8,32,power_law_1.01,0.058567681312561036
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,8,32,power_law_1.2,0.10294015884399414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,8,32,power_law_1.2,0.10697983741760253
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,8,32,power_law_1.2,0.10962944030761719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,8,32,power_law_1.2,0.11239551544189452
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,8,32,power_law_1.2,0.11259391784667969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,8,32,power_law_1.2,0.11306879997253419
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,8,32,power_law_1.2,0.11823616027832032
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,8,32,power_law_1.2,0.12988160133361817
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,8,32,power_law_1.2,0.0553433609008789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,8,32,power_law_1.2,0.14263039588928222
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,8,32,power_law_1.2,0.055592961311340325
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,8,32,power_law_1.2,0.18156543731689453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,8,32,power_law_1.2,0.20246015548706056
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,8,32,power_law_1.2,0.2739737510681152
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,8,32,power_law_1.2,0.3588825607299805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,8,32,power_law_1.2,0.03639679908752441
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,8,32,power_law_1.2,0.036898560523986816
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,8,32,power_law_1.2,0.03633536100387573
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,8,32,power_law_1.2,0.5158156967163086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,8,32,power_law_1.2,0.0375871992111206
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,8,32,power_law_1.2,0.05345920085906982
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,8,32,power_law_1.2,0.05442560195922852
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,8,32,power_law_1.2,0.6932838439941407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,8,32,power_law_1.2,0.05729152202606201
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,8,32,power_law_1.2,0.057876482009887695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,8,32,power_law_1.2,0.059102721214294426
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,8,32,power_law_1.2,0.9253043365478515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,8,32,power_law_1.2,0.06872831821441651
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,8,32,power_law_1.2,0.06630784034729004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,8,32,power_law_1.2,0.06488063812255859
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,8,32,power_law_1.2,0.0711359977722168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,8,32,power_law_1.2,0.07098879814147949
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,8,32,power_law_1.2,0.08567935943603515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,8,32,power_law_1.2,0.10262656211853027
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,8,32,power_law_1.2,0.13399552345275878
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,8,32,power_law_1.2,1.4407615661621094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,8,32,power_law_1.2,0.15446399688720702
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,8,32,power_law_1.2,0.20847232818603514
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,8,32,power_law_1.2,0.2749760055541992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,8,32,power_law_1.2,0.06025216102600097
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,8,32,power_law_1.2,0.4359539031982422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,16,16,balanced,0.02717184066772461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,8,32,power_law_1.2,0.5339737701416015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,16,16,balanced,0.027399680614471435
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,16,16,balanced,0.02824320077896118
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,8,32,power_law_1.2,0.07404287815093993
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,16,16,balanced,0.029562880992889406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,16,16,balanced,0.04865151882171631
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,16,16,balanced,0.0609984016418457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,16,16,balanced,0.09960063934326172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,16,16,balanced,0.09953408241271973
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,16,16,balanced,0.10528512001037597
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,16,16,balanced,0.10502016067504882
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,8,32,power_law_1.2,1.046725082397461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,16,16,balanced,0.1115187168121338
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,16,16,balanced,0.11429759979248047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,16,16,balanced,0.10710911750793457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,16,16,balanced,0.11124223709106444
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,16,16,balanced,0.11875455856323243
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,16,16,balanced,0.12737024307250977
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,8,32,power_law_1.2,0.06108543872833252
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,16,16,balanced,0.15373824119567872
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,16,16,balanced,0.1725632095336914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,16,16,balanced,0.2107353591918945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,8,32,power_law_1.2,0.7489395141601562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,16,16,balanced,0.24762239456176757
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,16,16,balanced,0.322369270324707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,16,16,balanced,0.10854656219482421
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,16,16,balanced,0.3973235321044922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,16,16,balanced,0.02783744096755981
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,16,16,balanced,0.027733759880065916
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,16,16,balanced,0.028142080307006833
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,16,16,balanced,0.5524851226806641
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,16,16,balanced,0.03291264057159424
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,16,16,balanced,0.735014419555664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,16,16,balanced,0.07211775779724121
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,16,16,balanced,0.06534912109374999
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,16,16,balanced,0.06826111793518067
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,16,16,balanced,0.06880767822265625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,16,16,balanced,0.0714188814163208
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,16,16,balanced,0.0767577600479126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,16,16,balanced,0.06926335811614989
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,16,16,balanced,0.07269504070281982
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,16,16,balanced,0.07500160217285157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,16,16,balanced,0.0795750379562378
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,16,16,balanced,0.0860159969329834
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,16,16,balanced,0.09643263816833496
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,16,16,balanced,0.11500800132751465
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,16,16,balanced,0.11276160240173341
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,16,16,balanced,0.029795839786529544
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,16,16,balanced,0.1325312042236328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,16,16,balanced,0.1687513542175293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,16,16,balanced,0.04289919853210449
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,16,16,balanced,0.2033075141906738
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,16,16,balanced,0.2853209686279297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,16,16,power_law_1.01,0.03926016092300415
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,16,16,power_law_1.01,0.056872959136962886
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,16,16,balanced,0.07406208038330078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,16,16,power_law_1.01,0.056556801795959476
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,16,16,power_law_1.01,0.05797247886657715
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,16,16,power_law_1.01,0.06130047798156738
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,16,16,balanced,0.6568000030517578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,16,16,power_law_1.01,0.09452032089233399
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,16,16,power_law_1.01,0.09659520149230957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,16,16,power_law_1.01,0.1036800003051758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,16,16,power_law_1.01,0.1051865577697754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,16,16,power_law_1.01,0.10790271759033203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,16,16,power_law_1.01,0.11220735549926757
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,16,16,power_law_1.01,0.11405183792114257
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,16,16,power_law_1.01,0.10811776161193846
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,16,16,power_law_1.01,0.11252736091613769
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,16,16,power_law_1.01,0.11676544189453124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,16,16,power_law_1.01,0.12111871719360352
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,16,16,balanced,0.352619514465332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,16,16,power_law_1.01,0.1351372814178467
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,16,16,power_law_1.01,0.14190336227416994
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,16,16,power_law_1.01,0.1676288032531738
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,16,16,power_law_1.01,0.19837440490722655
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,16,16,power_law_1.01,0.25323904037475586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,16,16,power_law_1.01,0.29521791458129887
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,16,16,power_law_1.01,0.10051327705383302
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,16,16,power_law_1.01,0.4516684722900391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,16,16,power_law_1.01,0.03216000080108643
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,16,16,power_law_1.01,0.03922688007354737
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,16,16,balanced,0.13537535667419434
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,16,16,power_law_1.01,0.5924915313720703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,16,16,power_law_1.01,0.036976640224456785
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,16,16,power_law_1.01,0.041118721961975094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,16,16,power_law_1.01,0.04227712154388428
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,16,16,power_law_1.01,0.06157567977905274
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,16,16,power_law_1.01,0.06372608184814453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,16,16,balanced,0.5059929656982421
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,16,16,power_law_1.01,0.06417151927947998
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,16,16,power_law_1.01,0.06608511924743651
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,16,16,power_law_1.01,0.06617983818054199
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,16,16,power_law_1.01,0.07177984237670898
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,16,16,power_law_1.01,0.07484288215637207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,16,16,power_law_1.01,0.07074048042297362
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,16,16,power_law_1.01,0.07370880126953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,16,16,power_law_1.01,0.07719679832458495
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,16,16,power_law_1.01,0.08010880470275879
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,16,16,power_law_1.01,0.09844736099243165
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,16,16,power_law_1.01,0.10737919807434082
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,16,16,power_law_1.01,0.13650688171386718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,16,16,power_law_1.01,0.17570560455322265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,16,16,power_law_1.01,0.2154470443725586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,16,16,power_law_1.01,0.2706777572631836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,16,16,power_law_1.01,0.3590924835205078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,16,16,power_law_1.01,0.4991462326049804
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,16,16,power_law_1.01,1.104949722290039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,16,16,power_law_1.2,0.03732608079910278
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,16,16,power_law_1.01,0.6773824310302734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,16,16,power_law_1.2,0.05607168197631836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,16,16,power_law_1.2,0.060417280197143555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,16,16,power_law_1.2,0.05787263870239258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,16,16,power_law_1.2,0.05978752136230468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,16,16,power_law_1.2,0.09516160011291504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,16,16,power_law_1.01,0.8974642944335937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,16,16,power_law_1.2,0.0986905574798584
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,16,16,power_law_1.2,0.0992460823059082
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,16,16,power_law_1.01,0.7770931243896484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,16,16,power_law_1.2,0.10196479797363281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,16,16,power_law_1.2,0.10368639945983886
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,16,16,power_law_1.2,0.11118720054626466
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,16,16,power_law_1.2,0.11127807617187499
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,16,16,power_law_1.2,0.11189375877380373
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,16,16,power_law_1.2,0.11122048377990723
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,16,16,power_law_1.2,0.11466879844665527
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,16,16,power_law_1.2,0.12041855812072755
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,16,16,power_law_1.2,0.13968128204345703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,16,16,power_law_1.2,0.1563046360015869
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,16,16,power_law_1.2,0.192040958404541
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,16,16,power_law_1.2,0.21375999450683594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,16,16,power_law_1.2,0.26269184112548827
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,16,16,power_law_1.2,0.3362112045288086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,16,16,power_law_1.2,0.5064102554321289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,16,16,power_law_1.2,0.032264959812164304
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,16,16,power_law_1.2,0.03913984060287475
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,16,16,power_law_1.2,0.6797376251220703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,16,16,power_law_1.2,0.0396940803527832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,16,16,power_law_1.2,0.11704704284667969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,16,16,power_law_1.01,0.07660927772521972
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,16,16,power_law_1.2,0.04238207817077637
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,16,16,power_law_1.2,0.06446080207824707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,16,16,power_law_1.2,0.06570496082305907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,16,16,power_law_1.2,0.064683518409729
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,16,16,power_law_1.2,0.06724607944488525
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,16,16,power_law_1.2,0.06719359874725342
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,16,16,power_law_1.2,1.0683443450927734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,16,16,power_law_1.2,0.06980864048004151
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,16,16,power_law_1.2,0.07357439994812012
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,16,16,power_law_1.2,0.07289599895477294
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,16,16,power_law_1.2,0.07542784214019775
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,16,16,power_law_1.2,0.07742464065551757
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,16,16,power_law_1.2,0.08108415603637695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,16,16,power_law_1.2,0.09542911529541016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,16,16,power_law_1.2,0.10865023612976074
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,16,16,power_law_1.2,0.14111743927001955
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,16,16,power_law_1.2,0.17063936233520507
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,16,16,power_law_1.2,0.041543679237365724
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,16,16,power_law_1.2,0.22015104293823243
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,16,16,power_law_1.2,0.28490495681762695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,32,8,balanced,0.027732479572296142
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,32,8,balanced,0.028520960807800293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,16,16,power_law_1.2,0.418389778137207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,32,8,balanced,0.031756799221038814
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,16,16,power_law_1.2,0.5094604873657227
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,16,16,power_law_1.2,0.07711872100830078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,32,8,balanced,0.06959487915039062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,32,8,balanced,0.1079155158996582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,32,8,balanced,0.11140864372253419
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,32,8,balanced,0.11459327697753907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,16,16,power_law_1.2,1.5012351989746093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,16,16,power_law_1.2,0.8667686462402344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,32,8,balanced,0.11685631752014161
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,32,8,balanced,0.10902655601501465
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,32,8,balanced,0.11198975563049315
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,32,8,balanced,0.11796735763549804
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,32,8,balanced,0.12222463607788085
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,32,8,balanced,0.11937408447265625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,16,16,power_law_1.2,1.0074317169189453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,32,8,balanced,0.135729923248291
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,32,8,balanced,0.1541312026977539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,32,8,balanced,0.17055871963500976
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,32,8,balanced,0.20640384674072268
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,32,8,balanced,0.24304128646850587
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,32,8,balanced,0.31611135482788083
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,32,8,balanced,0.05658239841461181
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,32,8,balanced,0.3873497772216797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,32,8,balanced,0.028569600582122802
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,32,8,balanced,0.5312358474731445
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,32,8,balanced,0.028974080085754396
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,32,8,balanced,0.031101438999176022
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,32,8,balanced,0.03459712028503418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,32,8,balanced,0.04030464172363281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,32,8,balanced,0.12842368125915526
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,32,8,balanced,0.08256768226623536
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,32,8,balanced,0.0850534439086914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,32,8,balanced,0.9693529510498047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,32,8,balanced,0.09031423568725586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,32,8,balanced,0.0858944034576416
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,32,8,balanced,0.038867199420928956
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,32,8,balanced,1.2945843505859376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,32,8,balanced,0.09137791633605957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,32,8,balanced,0.09348992347717286
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,32,8,balanced,0.09807871818542481
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,32,8,balanced,0.6772108459472657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,32,8,balanced,0.10895744323730469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,32,8,balanced,0.12643839836120604
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,32,8,balanced,0.0584332799911499
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,32,8,balanced,0.14398591995239257
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,32,8,balanced,0.08770688056945801
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,32,8,balanced,0.08241791725158691
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,32,8,balanced,0.2141823959350586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,32,8,balanced,0.2834265518188476
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,32,8,balanced,0.3492812728881836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,32,8,balanced,0.08789119720458985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,32,8,power_law_1.01,0.03017600059509277
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,32,8,balanced,0.5119551849365235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,32,8,power_law_1.01,0.04807424068450928
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,32,8,balanced,0.6477632141113281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,32,8,power_law_1.01,0.0561740779876709
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,32,8,power_law_1.01,0.05875840187072754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,32,8,power_law_1.01,0.06257919788360596
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,32,8,balanced,0.9461785888671874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,32,8,power_law_1.01,0.06322559833526611
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,32,8,balanced,0.17927936553955076
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,32,8,power_law_1.01,0.10702719688415527
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,32,8,power_law_1.01,0.10949888229370117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,32,8,power_law_1.01,0.11690112113952637
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,32,8,power_law_1.01,0.11179264068603516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,32,8,power_law_1.01,0.11054847717285157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,32,8,power_law_1.01,0.11643903732299805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,32,8,power_law_1.01,0.11566975593566894
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,32,8,power_law_1.01,0.11832192420959473
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,32,8,power_law_1.01,0.1207756805419922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,32,8,power_law_1.01,0.12759936332702637
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,32,8,power_law_1.01,0.13966079711914064
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,32,8,power_law_1.01,0.16172800064086915
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,32,8,power_law_1.01,0.19468799591064453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,32,8,power_law_1.01,0.21752576828002929
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,32,8,power_law_1.01,0.30436607360839846
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,32,8,power_law_1.01,0.3856563186645508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,32,8,power_law_1.01,0.10288000106811523
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,32,8,power_law_1.01,0.032308480739593505
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,32,8,power_law_1.01,0.4716287994384766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,32,8,power_law_1.01,0.03697792053222656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,32,8,balanced,1.247869415283203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,32,8,power_law_1.01,0.6220876693725585
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,32,8,power_law_1.01,0.04419583797454834
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,32,8,power_law_1.01,0.04609663963317871
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,32,8,power_law_1.01,0.052312321662902825
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,32,8,power_law_1.01,0.051104001998901374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,32,8,power_law_1.01,0.08039423942565918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,32,8,power_law_1.01,0.07938176155090332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,32,8,power_law_1.01,0.08423808097839355
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,32,8,power_law_1.01,0.9972249603271484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,32,8,power_law_1.01,0.08781184196472168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,32,8,power_law_1.01,0.08728704452514649
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,32,8,power_law_1.01,0.0813657569885254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,32,8,power_law_1.01,1.2164224243164061
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,32,8,power_law_1.01,0.0949888038635254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,32,8,power_law_1.01,0.09072768211364746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,32,8,power_law_1.01,0.1161574363708496
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,32,8,power_law_1.01,0.1366464042663574
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,32,8,power_law_1.01,0.17013248443603515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,32,8,power_law_1.01,0.20339839935302734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,32,8,power_law_1.01,0.27388160705566406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,32,8,power_law_1.01,0.3391027069091797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,32,8,power_law_1.01,0.07603328227996826
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,32,8,power_law_1.01,0.47625854492187497
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,32,8,power_law_1.2,0.03082495927810669
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,32,8,power_law_1.01,0.5870080184936524
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,32,8,power_law_1.01,0.08698623657226563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,32,8,power_law_1.2,0.047964158058166506
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,32,8,power_law_1.01,0.08792320251464844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,32,8,power_law_1.2,0.05302527904510498
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,32,8,power_law_1.2,0.10528767585754395
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,32,8,power_law_1.2,0.061181440353393554
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,32,8,power_law_1.2,0.0669811201095581
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,32,8,power_law_1.2,0.10488575935363768
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,32,8,power_law_1.01,0.9158886718750001
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,32,8,power_law_1.2,0.10988415718078612
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,32,8,power_law_1.2,0.11666175842285156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,32,8,power_law_1.2,0.11109248161315918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,32,8,power_law_1.2,0.11932800292968751
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,32,8,power_law_1.2,0.11403903961181641
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,32,8,power_law_1.01,1.2410034942626953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,32,8,power_law_1.2,0.12161664009094238
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,32,8,power_law_1.2,0.12868096351623534
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,32,8,power_law_1.2,0.15016192436218262
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,32,8,power_law_1.2,0.166561279296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,32,8,power_law_1.2,0.23388416290283204
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,32,8,power_law_1.2,0.05584383964538574
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,32,8,power_law_1.2,0.3051174354553223
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,32,8,power_law_1.2,0.3739878463745117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,32,8,power_law_1.2,0.5527910232543946
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,32,8,power_law_1.2,0.11019392013549804
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,32,8,power_law_1.2,0.11020799636840821
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,32,8,power_law_1.2,0.032522239685058595
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,32,8,power_law_1.2,0.0384768009185791
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,32,8,power_law_1.2,0.037242879867553716
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,32,8,power_law_1.2,0.04294655799865722
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,32,8,power_law_1.2,0.05015039920806885
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,32,8,power_law_1.2,1.0097369384765624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,32,8,power_law_1.2,0.07535488128662109
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,32,8,power_law_1.2,0.05619840145111084
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,32,8,power_law_1.2,0.07818880081176757
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,32,8,power_law_1.2,0.0827724838256836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,32,8,power_law_1.2,0.08753151893615722
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,32,8,power_law_1.2,1.295047607421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,32,8,power_law_1.2,0.0850163173675537
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,32,8,power_law_1.2,0.08659968376159669
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,32,8,power_law_1.2,0.09102848052978516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,32,8,power_law_1.2,0.6993574523925782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,32,8,power_law_1.2,0.10330240249633789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,32,8,power_law_1.2,0.11838080406188964
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,32,8,power_law_1.2,0.19782400131225586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,32,8,power_law_1.2,0.14079360008239744
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,32,8,power_law_1.2,0.1745792007446289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,32,8,power_law_1.2,0.2085990333557129
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,32,8,power_law_1.2,0.0825369644165039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,32,8,power_law_1.2,0.2878438377380371
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,32,8,power_law_1.2,0.3676313781738281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,32,8,power_law_1.2,0.08132863998413085
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,2,128,balanced,0.016307200193405154
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,32,8,power_law_1.2,0.5229926300048828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,32,8,power_law_1.2,0.09261183738708496
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,2,128,balanced,0.016590080261230468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,2,128,balanced,0.01637887954711914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,32,8,power_law_1.2,0.709381103515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,2,128,balanced,0.016373759508132933
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,2,128,balanced,0.022110719680786133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,2,128,balanced,0.02230655908584595
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,2,128,balanced,0.022577919960021973
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,2,128,balanced,0.02268671989440918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,2,128,balanced,0.023009281158447265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,2,128,balanced,0.023366398811340332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,2,128,balanced,0.02335616111755371
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,2,128,balanced,0.023878400325775144
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,2,128,balanced,0.025512959957122806
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,32,8,power_law_1.2,1.0455615997314454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,2,128,balanced,0.027392001152038575
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,2,128,balanced,0.029196801185607912
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,2,128,balanced,0.02789760112762451
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,2,128,balanced,0.029982719421386715
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,32,8,power_law_1.2,1.2237401580810547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,2,128,balanced,0.03197567939758301
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,2,128,balanced,0.0369651198387146
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,2,128,balanced,0.016401920318603515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,2,128,balanced,0.04958975791931152
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,2,128,balanced,0.07254655838012695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,2,128,balanced,0.019359999895095827
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,2,128,balanced,0.01930624008178711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,2,128,balanced,0.02236288070678711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,2,128,balanced,0.019174400568008423
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,2,128,balanced,0.028436479568481443
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,2,128,balanced,0.028733439445495605
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,2,128,balanced,0.026303999423980713
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,2,128,balanced,0.02868351936340332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,2,128,balanced,0.031805438995361326
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,2,128,balanced,0.02891648054122925
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,2,128,balanced,0.029239039421081543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,2,128,balanced,0.029551360607147213
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,2,128,balanced,0.029304320812225344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,2,128,balanced,0.029888000488281254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,2,128,balanced,0.019285759925842284
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,2,128,balanced,0.03043839931488037
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,2,128,balanced,0.031470079421997074
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,2,128,balanced,0.032222719192504884
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,2,128,balanced,0.03387135982513428
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,2,128,balanced,0.0367411208152771
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,2,128,balanced,0.03341439962387085
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,2,128,balanced,0.03693311929702759
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,2,128,balanced,0.025391359329223633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,2,128,balanced,0.0404966402053833
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,2,128,balanced,0.04649983882904053
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,2,128,balanced,0.029157121181488038
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,2,128,balanced,0.05711743831634521
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,2,128,balanced,0.06659711837768555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,2,128,power_law_1.01,0.016272640228271483
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,2,128,power_law_1.01,0.016244479417800904
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,2,128,power_law_1.01,0.016321280002593995
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,2,128,power_law_1.01,0.02212735891342163
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,2,128,balanced,0.06286975860595703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,2,128,power_law_1.01,0.022639360427856445
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,2,128,power_law_1.01,0.023002879619598387
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,2,128,power_law_1.01,0.023647999763488768
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,2,128,power_law_1.01,0.023760640621185304
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,2,128,power_law_1.01,0.024724481105804445
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,2,128,power_law_1.01,0.02535936117172241
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,2,128,power_law_1.01,0.02717695951461792
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,2,128,power_law_1.01,0.02841599941253662
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,2,128,power_law_1.01,0.03083775997161865
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,2,128,balanced,0.035092480182647705
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,2,128,power_law_1.01,0.03198335886001587
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,2,128,power_law_1.01,0.03190655946731567
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,2,128,power_law_1.01,0.030709760189056395
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,2,128,power_law_1.01,0.034088959693908696
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,2,128,power_law_1.01,0.04765439987182617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,2,128,power_law_1.01,0.05925504207611084
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,2,128,power_law_1.01,0.0701964807510376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,2,128,power_law_1.01,0.09812479972839355
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,2,128,power_law_1.01,0.12223487854003907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,2,128,power_law_1.01,0.18471040725708007
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,2,128,power_law_1.01,0.2645248031616211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,2,128,balanced,0.029999361038208005
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,2,128,power_law_1.01,0.01938815951347351
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,2,128,power_law_1.01,0.38829311370849606
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,2,128,power_law_1.01,0.01911679983139038
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,2,128,power_law_1.01,0.01916159987449646
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,2,128,power_law_1.01,0.027537920475006104
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,2,128,power_law_1.01,0.02867840051651001
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,2,128,power_law_1.01,0.02927488088607788
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,2,128,power_law_1.01,0.029364480972290042
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,2,128,power_law_1.01,0.03000447988510132
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,2,128,power_law_1.01,0.030092799663543703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,2,128,power_law_1.01,0.03207040071487427
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,2,128,power_law_1.01,0.031233279705047606
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,2,128,power_law_1.01,0.032147200107574464
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,2,128,power_law_1.01,0.033061120510101315
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,2,128,power_law_1.01,0.035050239562988285
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,2,128,power_law_1.01,0.0372544002532959
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,2,128,power_law_1.01,0.03710207939147949
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,2,128,power_law_1.01,0.034777600765228275
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,2,128,power_law_1.01,0.03823487997055054
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,2,128,power_law_1.01,0.04308479785919189
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,2,128,power_law_1.01,0.05219071865081787
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,2,128,power_law_1.01,0.06123519897460937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,2,128,power_law_1.01,0.08133376121520997
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,2,128,balanced,0.019215359687805175
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,2,128,power_law_1.01,0.105163516998291
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,2,128,power_law_1.01,0.15521023750305177
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,2,128,power_law_1.01,0.18553855895996094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,2,128,power_law_1.2,0.016414719820022582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,2,128,power_law_1.01,0.3150067138671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,2,128,power_law_1.2,0.022042880058288576
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,2,128,power_law_1.2,0.022287359237670896
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,2,128,power_law_1.01,0.41860992431640626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,2,128,power_law_1.2,0.023089919090270996
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,2,128,power_law_1.2,0.022654719352722168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,2,128,power_law_1.2,0.023466238975524904
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,2,128,power_law_1.2,0.024235520362854004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,2,128,power_law_1.2,0.026749439239501953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,2,128,power_law_1.2,0.026143999099731446
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,2,128,power_law_1.2,0.028830718994140626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,2,128,power_law_1.2,0.030699520111083983
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,2,128,power_law_1.2,0.03220864057540894
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,2,128,power_law_1.2,0.030803198814392092
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,2,128,power_law_1.01,0.4519091033935547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,2,128,power_law_1.2,0.030705919265747072
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,2,128,power_law_1.2,0.03365504026412964
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,2,128,power_law_1.2,0.05007743835449219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,2,128,power_law_1.2,0.058462719917297366
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,2,128,power_law_1.2,0.08334079742431641
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,2,128,power_law_1.2,0.09958527565002442
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,2,128,power_law_1.2,0.016165119409561158
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,2,128,power_law_1.2,0.14529279708862303
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,2,128,power_law_1.2,0.027251200675964354
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,2,128,power_law_1.2,0.019221760034561157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,2,128,power_law_1.01,0.028637440204620362
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,2,128,power_law_1.2,0.01933184027671814
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,2,128,power_law_1.2,0.02823040008544922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,2,128,power_law_1.2,0.19009023666381836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,2,128,power_law_1.2,0.028582398891448972
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,2,128,power_law_1.2,0.029076480865478517
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,2,128,power_law_1.2,0.029114880561828614
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,2,128,power_law_1.2,0.030174720287322997
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,2,128,power_law_1.2,0.030017280578613283
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,2,128,power_law_1.2,0.030443520545959474
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,2,128,power_law_1.2,0.03188607931137085
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,2,128,power_law_1.2,0.03338367938995361
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,2,128,power_law_1.2,0.035000319480896
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,2,128,power_law_1.2,0.03771647930145264
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,2,128,power_law_1.2,0.03644288063049316
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,2,128,power_law_1.2,0.034897921085357664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,2,128,power_law_1.2,0.03744127988815308
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,2,128,power_law_1.2,0.044686079025268555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,2,128,power_law_1.2,0.05111423969268799
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,2,128,power_law_1.2,0.06954880237579346
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,2,128,power_law_1.2,0.08740608215332031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,2,128,power_law_1.2,0.1266521644592285
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,2,128,power_law_1.2,0.16333440780639646
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,2,128,power_law_1.2,0.031093759536743166
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,2,128,power_law_1.2,0.3038899230957031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,4,64,balanced,0.016028159856796266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,2,128,power_law_1.2,0.24686208724975586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,4,64,balanced,0.01592192053794861
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,4,64,balanced,0.016031999588012696
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,4,64,balanced,0.016087039709091186
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,4,64,balanced,0.016024320125579833
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,4,64,balanced,0.02153088092803955
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,4,64,balanced,0.02189824104309082
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,4,64,balanced,0.022133760452270508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,4,64,balanced,0.022849280834198
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,4,64,balanced,0.022753279209136962
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,4,64,balanced,0.023462400436401368
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,4,64,balanced,0.023673601150512695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,4,64,balanced,0.024149758815765383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,4,64,balanced,0.02496256113052368
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,4,64,balanced,0.02512383937835693
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,4,64,balanced,0.025639679431915287
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,4,64,balanced,0.02781696081161499
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,4,64,balanced,0.030640640258789063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,2,128,power_law_1.2,0.3899609756469727
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,4,64,balanced,0.026928639411926268
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,4,64,balanced,0.029632000923156737
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,4,64,balanced,0.03092351913452148
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,4,64,balanced,0.034142720699310306
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,4,64,balanced,0.04632063865661621
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,4,64,balanced,0.05812479972839355
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,4,64,balanced,0.06768256187438965
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,4,64,balanced,0.0877836799621582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,4,64,balanced,0.10542719841003419
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,4,64,balanced,0.019251199960708617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,4,64,balanced,0.0192960000038147
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,4,64,balanced,0.019288320541381836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,4,64,balanced,0.01928063988685608
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,4,64,balanced,0.019502079486846922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,4,64,balanced,0.028354558944702145
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,4,64,balanced,0.028847360610961915
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,4,64,balanced,0.028770558834075928
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,4,64,balanced,0.029392640590667724
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,4,64,balanced,0.02910207986831665
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,4,64,balanced,0.029643518924713137
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,4,64,balanced,0.02980736017227173
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,4,64,balanced,0.03020031929016113
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,4,64,balanced,0.030958080291748048
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,4,64,balanced,0.031168000698089598
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,4,64,balanced,0.03160703897476196
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,4,64,balanced,0.033191680908203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,2,128,power_law_1.2,0.34169727325439453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,4,64,balanced,0.03195904016494751
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,4,64,balanced,0.03425407886505127
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,4,64,balanced,0.03634687900543213
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,4,64,balanced,0.03987200021743774
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,4,64,balanced,0.04405248165130615
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,4,64,balanced,0.05327744007110595
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,4,64,balanced,0.06341504096984864
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,4,64,balanced,0.08370944023132323
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,4,64,balanced,0.10482303619384765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,4,64,power_law_1.01,0.015909119844436645
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,4,64,power_law_1.01,0.01618559956550598
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,4,64,power_law_1.01,0.016284159421920776
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,4,64,power_law_1.01,0.02142080068588257
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,4,64,power_law_1.01,0.02210815906524658
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,4,64,power_law_1.01,0.022539520263671876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,4,64,balanced,0.035791358947753905
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,4,64,power_law_1.01,0.02322688102722168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,4,64,power_law_1.01,0.024463360309600828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,4,64,power_law_1.01,0.02491136074066162
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,4,64,power_law_1.01,0.024620800018310546
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,4,64,power_law_1.01,0.027310080528259277
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,4,64,power_law_1.01,0.02905600070953369
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,4,64,power_law_1.01,0.029418239593505858
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,4,64,power_law_1.01,0.028930559158325198
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,4,64,power_law_1.01,0.030220799446105957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,4,64,power_law_1.01,0.030624001026153563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,4,64,power_law_1.01,0.03913216114044189
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,4,64,power_law_1.01,0.04998784065246582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,4,64,power_law_1.01,0.058836479187011716
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,4,64,power_law_1.01,0.066048002243042
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,4,64,power_law_1.01,0.09620608329772949
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,4,64,power_law_1.01,0.11048064231872559
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,4,64,power_law_1.01,0.1711897659301758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,4,64,power_law_1.01,0.23687295913696288
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,4,64,power_law_1.01,0.02996992111206055
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,4,64,power_law_1.01,0.372542724609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,4,64,power_law_1.01,0.01936511993408203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,4,64,power_law_1.01,0.4363430404663086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,4,64,power_law_1.01,0.019554560184478757
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,4,64,power_law_1.01,0.019423999786376954
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,4,64,power_law_1.01,0.02824959993362427
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,4,64,power_law_1.01,0.028778240680694577
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,4,64,power_law_1.01,0.029018878936767578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,4,64,power_law_1.01,0.029561600685119628
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,4,64,power_law_1.01,0.030197761058807372
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,4,64,power_law_1.01,0.030529279708862305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,4,64,power_law_1.01,0.030557439327239987
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,4,64,power_law_1.01,0.03148927927017212
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,4,64,power_law_1.01,0.03287424087524414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,4,64,power_law_1.01,0.03377919912338257
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,4,64,power_law_1.01,0.03631103992462158
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,4,64,power_law_1.01,0.03475967884063721
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,4,64,power_law_1.01,0.03360383987426758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,4,64,power_law_1.01,0.034982399940490724
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,4,64,power_law_1.01,0.03816191911697388
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,4,64,power_law_1.01,0.04522111892700195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,4,64,power_law_1.01,0.05372799873352051
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,4,64,power_law_1.01,0.06998271942138672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,4,64,power_law_1.01,0.08799360275268556
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,4,64,power_law_1.01,0.1173363208770752
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,4,64,power_law_1.01,0.16046976089477538
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,4,64,power_law_1.01,0.21295103073120117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,4,64,power_law_1.01,0.2900313568115235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,4,64,power_law_1.2,0.01604096055030823
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,4,64,power_law_1.2,0.016213760375976563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,4,64,power_law_1.01,0.3953919982910156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,4,64,power_law_1.2,0.02142080068588257
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,4,64,power_law_1.2,0.0217574405670166
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,4,64,power_law_1.2,0.02255359888076782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,2,128,power_law_1.2,0.6558041381835937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,4,64,power_law_1.2,0.024042239189147947
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,4,64,power_law_1.2,0.02472575902938843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,4,64,power_law_1.2,0.024706559181213377
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,4,64,power_law_1.2,0.025649919509887698
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,4,64,power_law_1.2,0.02782464027404785
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,4,64,power_law_1.2,0.029241600036621095
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,4,64,power_law_1.2,0.029777920246124266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,4,64,power_law_1.2,0.028885760307312013
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,4,64,power_law_1.2,0.030393600463867188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,4,64,power_law_1.2,0.030280959606170655
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,4,64,power_law_1.2,0.03409791946411132
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,4,64,power_law_1.2,0.047655677795410155
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,4,64,power_law_1.2,0.05827583789825439
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,2,128,power_law_1.2,0.5083609771728516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,4,64,power_law_1.2,0.07546239852905273
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,4,64,power_law_1.2,0.09313023567199707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,4,64,power_law_1.2,0.022844159603118898
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,4,64,power_law_1.2,0.13362688064575196
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,4,64,power_law_1.2,0.14425215721130372
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,4,64,power_law_1.2,0.019188480377197264
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,4,64,power_law_1.2,0.019720959663391113
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,4,64,power_law_1.2,0.028204801082611086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,4,64,power_law_1.2,0.02871040105819702
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,4,64,power_law_1.2,0.029253120422363284
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,4,64,power_law_1.2,0.029282560348510744
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,4,64,power_law_1.2,0.03062527894973755
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,4,64,power_law_1.2,0.02995327949523926
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,4,64,power_law_1.2,0.03174272060394287
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,4,64,power_law_1.2,0.032517120838165284
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,4,64,power_law_1.2,0.25607295989990236
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,4,64,power_law_1.2,0.03347840070724487
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,2,128,power_law_1.2,0.6187136077880859
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,4,64,power_law_1.2,0.034663679599761965
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,4,64,power_law_1.2,0.035530240535736085
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,4,64,power_law_1.2,0.03517823934555054
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,4,64,power_law_1.2,0.03781248092651367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,4,64,power_law_1.2,0.0460595178604126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,4,64,power_law_1.2,0.03127935886383056
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,4,64,power_law_1.2,0.05123839855194092
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,4,64,power_law_1.2,0.07258368015289307
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,4,64,power_law_1.2,0.08033408164978027
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,4,64,power_law_1.2,0.1182592010498047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,4,64,power_law_1.2,0.03558271884918213
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,4,64,power_law_1.2,0.13813504219055175
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,4,64,power_law_1.2,0.3253209686279297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,2,128,balanced,0.023813118934631346
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,4,64,power_law_1.2,0.2169254493713379
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,2,128,balanced,0.023682560920715332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,2,128,balanced,0.023416318893432618
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,2,128,balanced,0.02349440097808838
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,2,128,balanced,0.02337536096572876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,2,128,balanced,0.035219199657440185
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,2,128,balanced,0.03495680093765259
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,2,128,balanced,0.035352320671081544
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,2,128,balanced,0.03558784008026123
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,2,128,balanced,0.03577215909957886
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,2,128,balanced,0.036038401126861574
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,2,128,balanced,0.036788480281829836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,2,128,balanced,0.037088000774383546
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,2,128,balanced,0.03779072046279907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,2,128,balanced,0.03846911907196045
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,2,128,balanced,0.03924992084503174
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,2,128,balanced,0.04473087787628174
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,2,128,balanced,0.0445030403137207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,2,128,balanced,0.04643839836120606
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,2,128,balanced,0.06097536087036133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,2,128,balanced,0.06518271923065186
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,2,128,balanced,0.07336448192596436
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,2,128,balanced,0.0822220802307129
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,2,128,balanced,0.09200384140014647
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,2,128,balanced,0.09884032249450683
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,2,128,balanced,0.12714624404907227
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,2,128,balanced,0.14127231597900392
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,2,128,balanced,0.024815359115600587
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,2,128,balanced,0.024714241027832033
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,2,128,balanced,0.024839680194854736
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,2,128,balanced,0.024826879501342776
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,4,64,power_law_1.2,0.3043315124511719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,2,128,balanced,0.02475519895553589
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,2,128,balanced,0.03957760095596313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,2,128,balanced,0.03976063966751099
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,2,128,balanced,0.03982464075088501
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,2,128,balanced,0.04001535892486573
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,2,128,balanced,0.040144639015197756
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,2,128,balanced,0.04028160095214844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,2,128,balanced,0.040609278678894044
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,2,128,balanced,0.040803837776184085
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,2,128,balanced,0.04135424137115479
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,2,128,balanced,0.04167808055877685
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,2,128,balanced,0.04225535869598389
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,2,128,balanced,0.043333120346069336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,2,128,balanced,0.04388735771179199
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,2,128,balanced,0.045735678672790527
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,2,128,balanced,0.04881408214569092
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,2,128,balanced,0.046743040084838865
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,2,128,balanced,0.05173120021820068
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,2,128,balanced,0.057815041542053226
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,2,128,balanced,0.06447487831115722
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,2,128,balanced,0.07181439876556397
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,2,128,balanced,0.08986751556396484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,2,128,balanced,0.10989824295043946
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,2,128,power_law_1.01,0.02346751928329468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,2,128,power_law_1.01,0.023394560813903807
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,2,128,power_law_1.01,0.023590400218963622
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,2,128,power_law_1.2,0.7957337951660156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,2,128,power_law_1.01,0.03513344049453736
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,2,128,power_law_1.01,0.03580928087234497
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,2,128,power_law_1.01,0.03636863946914673
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,2,128,power_law_1.01,0.036915199756622316
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,2,128,power_law_1.01,0.03783807992935181
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,2,128,power_law_1.01,0.039366400241851805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,2,128,power_law_1.01,0.041245441436767574
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,2,128,power_law_1.01,0.04594175815582276
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,2,128,power_law_1.01,0.04339456081390381
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,2,128,power_law_1.01,0.056606721878051755
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,2,128,power_law_1.01,0.06424448013305664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,4,64,power_law_1.2,0.418047981262207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,2,128,power_law_1.01,0.0715174388885498
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,2,128,power_law_1.01,0.07225599765777588
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,2,128,power_law_1.01,0.09190400123596192
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,2,128,power_law_1.01,0.08575615882873536
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,2,128,power_law_1.01,0.11947903633117676
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,2,128,power_law_1.01,0.14245375633239746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,2,128,power_law_1.01,0.1959052848815918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,2,128,power_law_1.01,0.24580736160278321
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,2,128,power_law_1.01,0.0433190393447876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,2,128,power_law_1.01,0.3276838302612305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,2,128,power_law_1.01,0.5180607986450195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,2,128,power_law_1.01,0.024747519493103026
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,2,128,power_law_1.01,0.024825599193572998
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,2,128,power_law_1.01,0.024613120555877686
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,2,128,power_law_1.01,0.03957632064819336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,2,128,power_law_1.01,0.04017663955688476
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,2,128,power_law_1.01,0.04086912155151367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,2,128,power_law_1.01,1.0379379272460938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,2,128,power_law_1.01,0.04136191844940186
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,2,128,power_law_1.01,0.0438259220123291
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,2,128,power_law_1.01,0.04293632030487061
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,2,128,power_law_1.01,0.04250751972198487
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,2,128,power_law_1.01,0.04388095855712891
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,2,128,power_law_1.01,0.04501632213592529
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,2,128,power_law_1.01,0.05105152130126953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,2,128,power_law_1.01,0.050864639282226565
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,2,128,power_law_1.01,0.7160205078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,2,128,power_law_1.01,0.053210878372192384
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,2,128,power_law_1.01,0.061100797653198244
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,2,128,power_law_1.01,0.040825600624084475
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,2,128,power_law_1.01,0.06507775783538819
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,2,128,power_law_1.01,0.08425984382629395
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,2,128,power_law_1.01,0.1007091236114502
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,2,128,power_law_1.01,0.1356595230102539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,2,128,power_law_1.01,0.1765977668762207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,2,128,power_law_1.01,0.2356211280822754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,2,128,power_law_1.01,0.04772863864898681
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,2,128,power_law_1.2,0.023303680419921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,2,128,power_law_1.01,0.3245849609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,2,128,power_law_1.2,0.02342655897140503
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,2,128,power_law_1.2,0.03527679920196533
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,2,128,power_law_1.01,0.49364353179931636
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,2,128,power_law_1.2,0.03542655944824219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,2,128,power_law_1.2,0.03583359956741333
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,2,128,power_law_1.2,0.036942079067230224
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,2,128,power_law_1.01,0.7041024017333984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,2,128,power_law_1.2,0.03811712026596069
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,2,128,power_law_1.2,0.03900671958923339
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,2,128,power_law_1.2,0.044604158401489256
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,2,128,power_law_1.2,0.043726081848144534
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,2,128,power_law_1.2,0.04315775871276856
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,2,128,power_law_1.2,0.05335936069488525
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,2,128,power_law_1.2,0.06774400234222412
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,2,128,power_law_1.2,0.07408768177032471
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,2,128,power_law_1.2,0.08398207664489746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,2,128,power_law_1.2,0.09447296142578125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,2,128,power_law_1.2,0.11720319747924805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,2,128,power_law_1.2,0.16515968322753907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,2,128,power_law_1.2,0.201977596282959
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,2,128,power_law_1.2,0.03683712005615235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,2,128,power_law_1.2,0.44516735076904296
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,2,128,power_law_1.2,0.3207411193847656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,2,128,power_law_1.2,0.07094016075134277
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,2,128,power_law_1.2,0.02495232105255127
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,2,128,power_law_1.2,0.02483583927154541
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,2,128,power_law_1.2,0.8564185333251952
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,2,128,power_law_1.2,0.039493119716644286
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,2,128,power_law_1.2,0.04015103816986084
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,2,128,power_law_1.2,0.040399360656738284
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,2,128,power_law_1.2,0.04106112003326416
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,2,128,power_law_1.2,0.040967679023742674
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,2,128,power_law_1.2,0.6831948852539063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,2,128,power_law_1.2,0.04347519874572754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,2,128,power_law_1.2,0.042158079147338864
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,2,128,power_law_1.2,0.0429747200012207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,2,128,power_law_1.2,0.043782401084899905
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,2,128,power_law_1.2,0.045401601791381835
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,2,128,power_law_1.2,0.04719488143920898
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,2,128,power_law_1.2,0.0511897611618042
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,2,128,power_law_1.2,0.05203584194183349
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,2,128,power_law_1.2,0.05298431873321533
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,2,128,power_law_1.2,0.05684351921081543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,2,128,power_law_1.2,0.06759679794311524
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,2,128,power_law_1.2,0.07996543884277343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,2,128,power_law_1.2,0.10636287689208986
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,2,128,power_law_1.2,0.1380288028717041
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,2,128,power_law_1.2,0.26801151275634766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,2,128,power_law_1.2,0.22816640853881837
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,4,64,power_law_1.2,0.49599998474121093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,2,128,power_law_1.2,1.4092416381835937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,4,64,balanced,0.02178431987762451
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,4,64,balanced,0.0216921591758728
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,4,64,balanced,0.02172032117843628
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,4,64,balanced,0.021333758831024167
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,4,64,balanced,0.022897920608520507
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,4,64,balanced,0.03304192066192627
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,4,64,balanced,0.033648641109466554
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,4,64,balanced,0.03366015911102295
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,2,128,power_law_1.2,0.5742156982421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,2,128,power_law_1.2,0.4149657440185547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,4,64,balanced,0.03528064012527466
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,4,64,balanced,0.0352128005027771
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,4,64,balanced,0.03751424074172974
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,4,64,balanced,0.03667200088500976
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,4,64,balanced,0.037191679477691655
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,4,64,balanced,0.04203519821166992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,4,64,balanced,0.03820287942886352
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,4,64,power_law_1.2,0.6717542266845703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,4,64,balanced,0.04135295867919922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,4,64,balanced,0.0602291202545166
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,4,64,balanced,0.06075263977050781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,4,64,balanced,0.06912255764007569
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,4,64,balanced,0.07526016235351562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,4,64,balanced,0.08404224395751954
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,4,64,balanced,0.09194496154785156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,4,64,balanced,0.1086079978942871
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,4,64,balanced,0.12245887756347655
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,4,64,balanced,0.1646758460998535
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,4,64,balanced,0.1956876754760742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,4,64,balanced,0.023607039451599122
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,4,64,balanced,0.023374080657958984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,4,64,balanced,0.023447039127349852
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,4,64,balanced,0.023532800674438477
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,4,64,balanced,0.02397439956665039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,4,64,balanced,0.0379801607131958
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,4,64,balanced,0.03854464054107666
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,4,64,balanced,0.039272959232330325
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,4,64,balanced,0.03879807949066162
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,4,64,balanced,0.03888511896133423
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,4,64,balanced,0.041118721961975094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,4,64,balanced,0.039855360984802246
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,4,64,balanced,0.04029056072235108
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,4,64,balanced,0.04473472118377685
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,4,64,balanced,0.04086656093597412
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,4,64,balanced,0.0416102409362793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,4,64,balanced,0.04321919918060303
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,4,64,balanced,0.04620416164398193
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,4,64,balanced,0.04472832202911377
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,4,64,balanced,0.049452800750732426
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,4,64,balanced,0.05459072113037109
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,4,64,balanced,0.05986815929412842
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,4,64,balanced,0.06636928081512451
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,4,64,balanced,0.08512639999389648
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,4,64,balanced,0.03834111928939819
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,4,64,balanced,0.09981311798095703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,4,64,balanced,0.1318617630004883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,4,64,balanced,0.16285568237304687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,4,64,power_law_1.01,0.022312960624694823
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,4,64,power_law_1.01,0.02243328094482422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,4,64,power_law_1.01,0.032875521183013914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,4,64,power_law_1.01,0.03362047910690308
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,4,64,power_law_1.01,0.035194880962371826
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,4,64,power_law_1.01,0.03565824031829834
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,4,64,power_law_1.01,0.03768959999084472
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,4,64,power_law_1.01,0.039002881050109864
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,4,64,power_law_1.01,0.03844223976135254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,4,64,power_law_1.01,0.040668158531188964
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,4,64,power_law_1.01,0.06046976089477539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,4,64,power_law_1.01,0.06387839794158937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,4,64,power_law_1.01,0.06677504062652588
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,4,64,power_law_1.01,0.02236032009124756
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,2,128,power_law_1.2,1.1490457916259766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,4,64,power_law_1.01,0.06918911933898926
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,4,64,power_law_1.01,0.07301631927490235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,4,64,power_law_1.01,0.08479871749877929
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,4,64,power_law_1.01,0.0926144027709961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,4,64,power_law_1.01,0.10456704139709472
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,4,64,power_law_1.01,0.12948863983154296
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,4,64,power_law_1.01,0.1713983917236328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,4,64,power_law_1.2,0.6032384109497071
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,4,64,power_law_1.01,0.23516159057617186
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,4,64,power_law_1.01,0.05160192012786865
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,4,64,power_law_1.01,0.34701183319091794
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,4,64,power_law_1.01,0.023775999546051023
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,4,64,power_law_1.01,0.023968000411987305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,4,64,power_law_1.01,0.023829760551452635
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,4,64,power_law_1.01,0.5381798553466797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,4,64,power_law_1.01,0.037876479625701905
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,4,64,power_law_1.01,0.03925503969192505
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,4,64,power_law_1.01,0.6958592224121094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,4,64,power_law_1.01,0.03899391889572144
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,4,64,power_law_1.01,0.039127039909362796
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,4,64,power_law_1.01,0.04041855812072754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,4,64,power_law_1.01,0.04212224006652832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,4,64,power_law_1.01,0.040711679458618165
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,4,64,power_law_1.01,0.04165120124816894
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,4,64,power_law_1.01,0.9188531494140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,4,64,power_law_1.01,0.0432371187210083
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,4,64,power_law_1.01,0.04488448143005371
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,4,64,power_law_1.01,0.04791296005249023
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,4,64,power_law_1.01,0.049539837837219235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,4,64,power_law_1.01,0.0509606409072876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,4,64,power_law_1.01,0.05335423946380615
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,4,64,power_law_1.01,0.059723520278930665
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,4,64,power_law_1.01,0.0677849578857422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,4,64,power_law_1.01,0.08729215621948241
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,4,64,power_law_1.01,0.11141759872436523
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,4,64,power_law_1.01,0.13715328216552733
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,2,128,power_law_1.2,1.8940966796874998
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,4,64,power_law_1.01,0.17395072937011719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,4,64,power_law_1.01,0.24831104278564453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,4,64,power_law_1.2,0.022664959430694583
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,4,64,power_law_1.2,0.02248447895050049
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,4,64,power_law_1.2,0.03277695894241333
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,4,64,power_law_1.01,0.5039795303344727
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,4,64,power_law_1.2,0.033456640243530275
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,4,64,power_law_1.2,0.0347980809211731
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,4,64,power_law_1.2,0.03515775918960572
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,4,64,power_law_1.2,0.0412992000579834
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,4,64,power_law_1.2,0.03594880104064942
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,4,64,power_law_1.2,0.03870719909667969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,4,64,power_law_1.2,0.05042943954467773
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,4,64,power_law_1.2,0.050405120849609374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,4,64,power_law_1.2,0.05814784049987794
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,4,64,power_law_1.2,0.0653708791732788
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,4,64,power_law_1.2,0.0730790376663208
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,4,64,power_law_1.01,0.3662643051147461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,4,64,power_law_1.2,0.09471487998962402
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,4,64,power_law_1.2,0.10472191810607909
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,4,64,power_law_1.2,0.14776191711425782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,4,64,power_law_1.01,0.6708223724365234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,4,64,power_law_1.2,0.16702592849731446
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,4,64,power_law_1.2,0.24554367065429689
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,4,64,power_law_1.2,0.29171712875366207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,4,64,power_law_1.2,0.07211904048919678
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,4,64,power_law_1.2,0.07916287899017334
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,4,64,power_law_1.2,0.023386878967285158
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,4,64,power_law_1.2,0.02403712034225464
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,4,64,power_law_1.2,0.03781120061874389
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,4,64,power_law_1.2,0.03937664031982422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,4,64,power_law_1.2,0.038602240085601806
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,4,64,power_law_1.2,0.038566401004791265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,4,64,power_law_1.2,0.6528179168701171
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,4,64,power_law_1.2,0.039677441120147705
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,4,64,power_law_1.2,0.041335039138793946
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,4,64,power_law_1.2,0.039792640209197996
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,4,64,power_law_1.2,0.04107391834259033
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,4,64,power_law_1.2,0.04141056060791016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,4,64,power_law_1.2,0.04313216209411621
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,4,64,power_law_1.2,0.04491136074066162
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,4,64,power_law_1.2,0.04592639923095703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,4,64,power_law_1.2,0.050586881637573244
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,4,64,power_law_1.2,0.052311038970947264
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,4,64,power_law_1.2,0.05578239917755127
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,4,64,power_law_1.2,0.06578048229217529
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,4,64,power_law_1.2,0.07937024116516114
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,4,64,power_law_1.2,0.09871999740600586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,4,64,power_law_1.2,0.1316710376739502
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,4,64,power_law_1.2,0.23868799209594727
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,4,64,power_law_1.2,0.19976831436157227
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,4,64,power_law_1.2,0.050122241973876956
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,2,128,power_law_1.2,0.8962598419189453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,4,64,power_law_1.2,0.48121726989746094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,8,32,balanced,0.021646080017089845
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,8,32,balanced,0.02145024061203003
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,8,32,balanced,0.021240320205688477
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,8,32,balanced,0.021841919422149657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,8,32,balanced,0.0232588791847229
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,8,32,balanced,0.033173758983612064
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,8,32,balanced,0.034792959690093994
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,8,32,balanced,0.03524991989135742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,8,32,balanced,0.03532799959182739
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,8,32,balanced,0.03572223901748657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,8,32,balanced,0.03725824117660522
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,8,32,balanced,0.03742208003997803
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,8,32,balanced,0.038380799293518064
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,8,32,balanced,0.040330238342285156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,8,32,balanced,0.04646399974822998
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,8,32,balanced,0.05493887901306153
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,8,32,balanced,0.05738368034362793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,8,32,balanced,0.06617728233337403
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,8,32,balanced,0.07250559806823731
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,8,32,balanced,0.08125696182250977
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,8,32,balanced,0.08822784423828126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,8,32,balanced,0.1038259220123291
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,8,32,balanced,0.11669504165649414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,8,32,balanced,0.14799103736877442
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,8,32,balanced,0.17752960205078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,8,32,balanced,0.23760768890380862
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,4,64,power_law_1.2,0.36757118225097657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,8,32,balanced,0.29827327728271485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,4,64,power_law_1.2,0.9533465576171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,8,32,balanced,0.023480319976806642
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,8,32,balanced,0.023452160358428956
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,8,32,balanced,0.02382591962814331
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,8,32,balanced,0.02508415937423706
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,8,32,balanced,0.03854207992553711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,8,32,balanced,0.0386572790145874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,8,32,balanced,0.039417600631713866
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,8,32,balanced,0.03971456050872803
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,8,32,balanced,0.04032639980316162
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,8,32,balanced,0.041070079803466795
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,8,32,balanced,0.04093567848205566
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,8,32,balanced,0.023339519500732424
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,4,64,power_law_1.2,1.0247782135009766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,8,32,balanced,0.04124927997589111
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,8,32,balanced,0.04314239978790283
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,8,32,balanced,0.04526080131530762
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,8,32,balanced,0.04445824146270752
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,8,32,balanced,0.04769408226013184
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,8,32,balanced,0.0496665620803833
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,8,32,balanced,0.053626880645751954
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,8,32,balanced,0.05885056018829345
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,8,32,balanced,0.06579455852508545
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,4,64,power_law_1.2,0.5382387161254882
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,8,32,balanced,0.08317312240600586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,8,32,balanced,0.09638400077819824
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,8,32,balanced,0.1261171245574951
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,8,32,balanced,0.15536767959594727
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,8,32,power_law_1.01,0.02234623908996582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,8,32,balanced,0.2141529655456543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,8,32,balanced,0.2703654479980469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,8,32,power_law_1.01,0.022690560817718506
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,8,32,power_law_1.01,0.023152639865875242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,8,32,power_law_1.01,0.033137919902801516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,8,32,power_law_1.01,0.03427072048187256
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,8,32,power_law_1.01,0.03557120084762573
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,8,32,power_law_1.01,0.03792511940002442
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,8,32,power_law_1.01,0.03739775896072388
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,8,32,power_law_1.01,0.04177408218383789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,8,32,power_law_1.01,0.059166722297668464
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,8,32,power_law_1.01,0.061223678588867195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,8,32,power_law_1.01,0.06444416046142579
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,8,32,power_law_1.01,0.06447999954223634
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,8,32,power_law_1.01,0.06786816120147705
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,8,32,power_law_1.01,0.07088255882263184
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,8,32,power_law_1.01,0.07336319923400879
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,8,32,power_law_1.01,0.08520575523376464
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,8,32,power_law_1.01,0.09099519729614258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,8,32,power_law_1.01,0.1170841598510742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,8,32,power_law_1.01,0.13812992095947266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,8,32,power_law_1.01,0.1849228858947754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,8,32,power_law_1.01,0.20822656631469724
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,8,32,power_law_1.01,0.29152511596679687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,8,32,power_law_1.01,0.03547647953033447
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,8,32,power_law_1.01,0.024687359333038332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,8,32,power_law_1.01,0.43208065032958987
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,8,32,power_law_1.01,0.02503551959991455
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,8,32,power_law_1.01,0.025242879390716556
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,8,32,power_law_1.01,0.037986559867858885
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,8,32,power_law_1.01,0.5524288177490234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,8,32,power_law_1.01,0.03953151941299439
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,8,32,power_law_1.01,0.039296000003814696
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,8,32,power_law_1.01,0.04032512187957764
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,8,32,power_law_1.01,0.040887041091918944
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,8,32,power_law_1.01,0.04107391834259033
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,8,32,power_law_1.01,0.04138239860534668
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,8,32,power_law_1.01,0.7874816131591797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,8,32,power_law_1.01,0.0427507209777832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,8,32,power_law_1.01,0.044545278549194336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,8,32,power_law_1.01,0.0478220796585083
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,8,32,power_law_1.01,0.048586239814758306
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,8,32,power_law_1.01,0.04994688034057617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,8,32,power_law_1.01,0.05339136123657227
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,8,32,power_law_1.01,0.06154240131378173
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,8,32,power_law_1.01,0.0735091209411621
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,8,32,power_law_1.01,0.08467455863952636
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,8,32,power_law_1.01,0.11569536209106446
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,8,32,power_law_1.01,0.14720383644104004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,8,32,power_law_1.01,0.18675071716308594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,8,32,power_law_1.01,0.25631616592407225
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,8,32,power_law_1.2,0.02261888027191162
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,8,32,power_law_1.2,0.022792959213256837
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,8,32,power_law_1.2,0.022787840366363527
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,8,32,power_law_1.01,0.4944793701171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,8,32,power_law_1.2,0.03286783933639527
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,8,32,power_law_1.2,0.0346342396736145
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,8,32,power_law_1.01,0.7218470764160156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,8,32,power_law_1.2,0.0354585599899292
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,8,32,power_law_1.2,0.036369919776916504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,8,32,power_law_1.2,0.037498879432678225
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,8,32,power_law_1.2,0.0400435209274292
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,8,32,power_law_1.2,0.05670015811920166
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,8,32,power_law_1.2,0.06118527889251709
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,8,32,power_law_1.2,0.06522624015808105
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,8,32,power_law_1.2,0.06701824188232422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,8,32,power_law_1.01,0.32479873657226566
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,8,32,power_law_1.2,0.07292160034179687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,8,32,power_law_1.2,0.08012672424316405
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,8,32,power_law_1.2,0.0896883201599121
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,8,32,power_law_1.2,0.09122431755065918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,8,32,power_law_1.2,0.1329868793487549
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,8,32,power_law_1.01,0.053052158355712885
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,8,32,power_law_1.2,0.14890368461608888
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,8,32,power_law_1.2,0.21601024627685547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,8,32,power_law_1.2,0.041150717735290526
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,8,32,power_law_1.2,0.06936063766479492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,8,32,power_law_1.2,0.28009216308593754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,8,32,power_law_1.2,0.0248524808883667
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,8,32,power_law_1.2,0.02499072074890137
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,8,32,power_law_1.2,0.42291713714599605
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,8,32,power_law_1.2,0.02480128049850464
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,8,32,power_law_1.2,0.03832319974899292
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,8,32,power_law_1.2,0.03880959987640381
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,8,32,power_law_1.2,0.039828479290008545
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,8,32,power_law_1.2,0.04274816036224365
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,8,32,power_law_1.2,0.04067967891693115
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,8,32,power_law_1.2,0.041324801445007324
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,8,32,power_law_1.2,0.041498880386352535
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,8,32,power_law_1.2,0.04495488166809082
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,8,32,power_law_1.2,0.04778111934661865
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,4,64,power_law_1.2,0.7687718200683593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,8,32,power_law_1.2,0.049927678108215336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,8,32,power_law_1.2,0.053473281860351565
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,8,32,power_law_1.2,0.05533184051513672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,8,32,power_law_1.2,0.06008959770202636
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,8,32,power_law_1.2,0.06831103801727295
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,8,32,power_law_1.2,0.07345536231994629
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,8,32,power_law_1.2,0.10379648208618164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,8,32,power_law_1.2,0.13005311965942384
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,8,32,power_law_1.2,0.17710079193115233
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,8,32,power_law_1.2,0.04769536018371582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,8,32,power_law_1.2,0.5859993743896484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,8,32,power_law_1.2,0.868532485961914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,16,16,balanced,0.020989439487457275
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,16,16,balanced,0.021523199081420898
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,16,16,balanced,0.02213759899139404
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,16,16,balanced,0.02309760093688965
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,8,32,power_law_1.2,0.24542976379394532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,16,16,balanced,0.025675520896911618
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,16,16,balanced,0.035852799415588374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,16,16,balanced,0.036728320121765134
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,16,16,balanced,0.038073599338531494
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,16,16,balanced,0.039118080139160155
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,16,16,balanced,0.04000383853912354
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,16,16,balanced,0.04588672161102295
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,16,16,balanced,0.05561855792999267
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,16,16,balanced,0.05536255836486816
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,16,16,balanced,0.06144000053405761
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,16,16,balanced,0.06549376010894775
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,16,16,balanced,0.03786623954772949
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,8,32,power_law_1.2,0.36777854919433595
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,16,16,balanced,0.07226367950439452
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,16,16,balanced,0.07930111885070801
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,16,16,balanced,0.08682239532470704
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,16,16,balanced,0.10051839828491212
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,16,16,balanced,0.11439104080200195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,16,16,balanced,0.14339839935302734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,16,16,balanced,0.1709657669067383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,16,16,balanced,0.22806400299072266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,16,16,balanced,0.2822336006164551
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,16,16,balanced,0.3910732650756836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,16,16,balanced,0.023301119804382327
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,16,16,balanced,0.02335616111755371
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,16,16,balanced,0.023756799697875978
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,16,16,balanced,0.023914239406585693
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,16,16,balanced,0.02596735954284668
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,16,16,balanced,0.040840959548950194
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,16,16,balanced,0.042997760772705076
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,16,16,balanced,0.04415743827819824
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,16,16,balanced,0.04337152004241943
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,16,16,balanced,0.04486400127410889
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,16,16,balanced,0.046314239501953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,8,32,power_law_1.2,0.7136409759521485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,16,16,balanced,0.048256001472473144
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,16,16,balanced,0.049374718666076664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,16,16,balanced,0.04815231800079346
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,16,16,balanced,0.04697472095489502
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,16,16,balanced,0.05391615867614746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,16,16,balanced,0.05871232032775879
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,8,32,power_law_1.2,0.5074150466918945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,16,16,balanced,0.06500607967376709
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,16,16,balanced,0.08118144035339356
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,16,16,balanced,0.09482751846313477
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,16,16,balanced,0.1247590446472168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,16,16,balanced,0.15111424446105956
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,16,16,balanced,0.20725631713867188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,16,16,balanced,0.26091007232666014
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,16,16,balanced,0.3737139129638672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,16,16,balanced,0.5028211212158202
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,16,16,power_law_1.01,0.023695359230041506
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,16,16,power_law_1.01,0.023935999870300293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,16,16,balanced,0.4806784057617188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,16,16,power_law_1.01,0.02392575979232788
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,16,16,power_law_1.01,0.025228800773620608
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,16,16,power_law_1.01,0.03635967969894409
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,16,16,power_law_1.01,0.036031999588012696
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,16,16,power_law_1.01,0.04039167881011963
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,16,16,power_law_1.01,0.04099584102630615
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,16,16,power_law_1.01,0.05090816020965576
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,16,16,power_law_1.01,0.0602239990234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,16,16,power_law_1.01,0.063787522315979
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,16,16,power_law_1.01,0.06440063953399658
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,16,16,power_law_1.01,0.06564735889434814
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,16,16,balanced,0.05035391807556152
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,16,16,power_law_1.01,0.07016191959381105
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,16,16,power_law_1.01,0.07419904232025146
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,16,16,power_law_1.01,0.07872128009796142
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,16,16,power_law_1.01,0.0890073585510254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,16,16,power_law_1.01,0.09302528381347656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,16,16,power_law_1.01,0.037911040782928465
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,16,16,power_law_1.01,0.1236736011505127
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,16,16,power_law_1.01,0.1443571186065674
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,16,16,power_law_1.01,0.18624895095825195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,8,32,power_law_1.2,1.1902681732177736
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,16,16,power_law_1.01,0.2424448013305664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,16,16,power_law_1.01,0.330832633972168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,16,16,power_law_1.01,0.41986560821533203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,16,16,power_law_1.01,0.024744958877563478
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,16,16,power_law_1.01,0.02499072074890137
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,16,16,power_law_1.01,0.025585920810699464
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,16,16,power_law_1.01,0.6373644638061523
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,16,16,power_law_1.01,0.039402239322662354
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,16,16,power_law_1.01,0.04124800205230713
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,16,16,power_law_1.01,0.0458355188369751
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,16,16,power_law_1.01,0.043956480026245116
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,16,16,power_law_1.01,0.8770342254638672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,16,16,power_law_1.01,0.045180158615112306
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,16,16,power_law_1.01,0.04548863887786865
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,16,16,power_law_1.01,0.04788224220275879
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,16,16,power_law_1.01,0.05155200004577637
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,16,16,power_law_1.01,0.049049601554870606
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,16,16,power_law_1.01,0.0533465576171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,16,16,power_law_1.01,0.054996480941772466
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,16,16,power_law_1.01,0.059877119064331054
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,16,16,power_law_1.01,0.07015935897827148
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,16,16,power_law_1.01,0.025136640071868898
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,16,16,power_law_1.01,0.07572224140167236
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,16,16,power_law_1.01,0.10292480468750001
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,16,16,power_law_1.01,0.12499456405639649
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,16,16,power_law_1.01,0.1575692844390869
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,16,16,power_law_1.01,0.20918527603149414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,16,16,power_law_1.01,0.3155737686157226
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,16,16,power_law_1.2,0.02367487907409668
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,16,16,power_law_1.01,0.3756915283203125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,16,16,power_law_1.01,0.05163263797760009
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,16,16,power_law_1.2,0.023831040859222413
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,16,16,power_law_1.2,0.024034559726715088
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,16,16,power_law_1.2,0.02557823896408081
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,16,16,power_law_1.2,0.03493247985839844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,16,16,power_law_1.01,0.5900019073486329
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,16,16,power_law_1.2,0.037556478977203364
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,16,16,power_law_1.2,0.037757439613342283
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,16,16,power_law_1.2,0.0387007999420166
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,16,16,power_law_1.01,0.7052339172363281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,4,64,power_law_1.2,1.3425689697265626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,16,16,power_law_1.2,0.04153855800628662
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,16,16,power_law_1.2,0.05911424160003662
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,16,16,power_law_1.2,0.06577407836914062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,16,16,power_law_1.2,0.06594048023223878
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,16,16,power_law_1.2,0.06904064178466797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,16,16,power_law_1.2,0.07344768047332764
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,16,16,power_law_1.2,0.07497600078582764
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,16,16,power_law_1.2,0.07990143775939942
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,16,16,power_law_1.2,0.10737407684326172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,8,32,power_law_1.2,1.0250816345214844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,16,16,power_law_1.2,0.13242112159729005
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,16,16,power_law_1.2,0.15925503730773927
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,16,16,power_law_1.2,0.20774911880493163
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,16,16,power_law_1.2,0.03612031936645508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,16,16,power_law_1.2,0.26336896896362305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,16,16,power_law_1.2,0.024737279415130615
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,16,16,power_law_1.2,0.025080320835113523
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,16,16,power_law_1.2,0.025414400100708008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,16,16,power_law_1.2,0.0893619155883789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,16,16,power_law_1.2,0.03978496074676514
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,16,16,power_law_1.2,0.04118912220001221
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,16,16,power_law_1.2,0.04322303771972656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,16,16,power_law_1.2,0.044153599739074706
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,16,16,power_law_1.2,0.04509568214416504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,16,16,power_law_1.2,0.04642687797546387
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,16,16,power_law_1.2,0.04971776008605957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,16,16,power_law_1.2,0.04874495983123779
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,16,16,power_law_1.2,0.38735614776611327
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,16,16,power_law_1.2,0.024764161109924316
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,16,16,power_law_1.2,0.05211904048919678
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,16,16,power_law_1.2,0.054712319374084474
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,16,16,power_law_1.2,0.05673727989196777
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,16,16,power_law_1.2,0.06105728149414062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,16,16,power_law_1.2,0.07026048183441162
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,16,16,power_law_1.2,0.08826751708984375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,16,16,power_law_1.2,0.11120896339416504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,16,16,power_law_1.2,0.1410764789581299
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,16,16,power_law_1.2,0.19371776580810546
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,16,16,power_law_1.2,0.23768447875976562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,16,16,power_law_1.2,0.5391052627563477
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,16,16,power_law_1.2,0.04887167930603027
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,32,8,balanced,0.021130239963531493
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,32,8,balanced,0.02215424060821533
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,32,8,balanced,0.02341248035430908
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,32,8,balanced,0.025593600273132323
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,32,8,balanced,0.029184000492095945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,32,8,balanced,0.04027647972106933
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,32,8,balanced,0.0490342378616333
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,16,16,power_law_1.2,0.733597412109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,32,8,balanced,0.05320703983306885
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,32,8,balanced,0.04908160209655762
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,32,8,balanced,0.06296703815460206
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,32,8,balanced,0.0550387191772461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,32,8,balanced,0.06458879947662353
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,32,8,balanced,0.0643558406829834
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,32,8,balanced,0.07260543823242187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,32,8,balanced,0.07554944038391113
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,32,8,balanced,0.07870463848114014
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,32,8,balanced,0.08523008346557617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,32,8,balanced,0.10124287605285645
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,32,8,balanced,0.11384703636169433
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,32,8,balanced,0.1410700798034668
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,32,8,balanced,0.16913536071777344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,16,16,power_law_1.2,0.5029145431518555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,32,8,balanced,0.22414335250854492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,16,16,power_law_1.2,0.3552012634277344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,32,8,balanced,0.27814144134521485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,32,8,balanced,0.3847014236450195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,32,8,balanced,0.4911884689331055
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,32,8,balanced,0.023809280395507813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,32,8,balanced,0.7097689819335937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,32,8,balanced,0.02407808065414429
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,32,8,balanced,0.024542720317840577
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,32,8,balanced,0.026202878952026366
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,32,8,balanced,0.04944896221160889
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,32,8,balanced,0.0510041618347168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,32,8,balanced,0.939557113647461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,32,8,balanced,0.052716798782348626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,32,8,balanced,0.05441152095794678
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,32,8,balanced,0.05060991764068603
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,32,8,balanced,0.052561922073364256
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,32,8,balanced,0.053315839767456054
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,32,8,balanced,0.05878143787384034
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,32,8,balanced,0.060117759704589845
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,32,8,balanced,0.061918721199035645
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,32,8,balanced,0.06938496112823486
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,32,8,balanced,0.08493184089660645
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,32,8,balanced,0.09873279571533203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,32,8,balanced,0.1251980781555176
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,32,8,balanced,0.15120384216308594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,32,8,balanced,0.030512640476226805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,32,8,balanced,0.25876352310180667
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,32,8,balanced,0.3717465591430664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,32,8,balanced,0.4772492980957031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,32,8,power_law_1.01,0.0232857608795166
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,32,8,balanced,0.696052474975586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,32,8,power_law_1.01,0.024190719127655032
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,32,8,balanced,0.048864002227783206
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,32,8,balanced,0.9168678283691406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,32,8,power_law_1.01,0.026216959953308104
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,32,8,power_law_1.01,0.02789760112762451
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,32,8,power_law_1.01,0.039815680980682375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,32,8,power_law_1.01,0.039701759815216064
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,32,8,power_law_1.01,0.04074751853942871
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,32,8,power_law_1.01,0.04150656223297119
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,32,8,power_law_1.01,0.060038399696350095
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,32,8,power_law_1.01,0.05869184017181397
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,32,8,power_law_1.01,0.06383615970611572
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,32,8,power_law_1.01,0.06572159767150879
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,32,8,power_law_1.01,0.07004032135009766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,32,8,power_law_1.01,0.07417215824127196
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,32,8,power_law_1.01,0.07683072090148926
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,32,8,power_law_1.01,0.08313216209411621
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,32,8,power_law_1.01,0.025108480453491212
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,32,8,power_law_1.01,0.09596416473388672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,32,8,power_law_1.01,0.10995840072631835
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,32,8,power_law_1.01,0.13200127601623535
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,32,8,power_law_1.01,0.16246784210205079
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,32,8,power_law_1.01,0.22137216567993162
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,32,8,power_law_1.01,0.26520576477050783
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,32,8,power_law_1.01,0.32661888122558597
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,32,8,power_law_1.01,0.43959678649902345
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,32,8,power_law_1.01,0.02518784046173096
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,32,8,balanced,0.21092351913452148
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,32,8,power_law_1.01,0.026978559494018554
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,32,8,power_law_1.01,0.027752959728240968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,32,8,power_law_1.01,0.7159884643554688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,32,8,power_law_1.01,0.028954880237579344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,32,8,power_law_1.01,0.02976639986038208
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,32,8,power_law_1.01,0.046913280487060546
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,32,8,power_law_1.01,0.047848958969116215
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,32,8,power_law_1.01,0.04887551784515381
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,32,8,power_law_1.01,0.051198720932006836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,32,8,power_law_1.01,0.9633446502685548
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,32,8,power_law_1.01,0.052645120620727535
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,32,8,power_law_1.01,0.05355264186859131
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,32,8,power_law_1.01,0.05396607875823974
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,32,8,power_law_1.01,0.05708928108215332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,32,8,power_law_1.01,0.05888895988464356
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,32,8,power_law_1.01,0.059644160270690916
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,32,8,power_law_1.01,0.06544511795043946
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,32,8,power_law_1.01,0.08131584167480468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,32,8,power_law_1.01,0.09514495849609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,32,8,power_law_1.01,0.12439552307128907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,32,8,power_law_1.01,0.15300992012023926
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,16,16,power_law_1.2,0.742371826171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,32,8,power_law_1.01,0.20291072845458985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,32,8,power_law_1.01,0.25045248031616213
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,32,8,power_law_1.01,0.374837760925293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,32,8,power_law_1.01,0.052330241203308106
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,32,8,power_law_1.01,0.49040382385253906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,32,8,power_law_1.2,0.02322943925857544
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,32,8,power_law_1.2,0.024294400215148927
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,32,8,power_law_1.2,0.02531455993652344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,32,8,power_law_1.01,0.6992639923095704
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,32,8,power_law_1.2,0.026003201007843018
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,32,8,power_law_1.2,0.027746560573577883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,16,16,power_law_1.2,1.0541209411621093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,32,8,power_law_1.2,0.047886080741882324
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,32,8,power_law_1.2,0.039891200065612795
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,32,8,power_law_1.2,0.049697279930114746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,32,8,power_law_1.01,0.9900262451171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,32,8,power_law_1.2,0.053738241195678715
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,32,8,power_law_1.2,0.06212736129760742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,32,8,power_law_1.2,0.06448639869689941
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,32,8,power_law_1.2,0.06363776206970215
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,32,8,power_law_1.2,0.07187839984893798
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,32,8,power_law_1.2,0.0772979211807251
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,32,8,power_law_1.2,0.08363648414611816
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,32,8,power_law_1.2,0.09577088356018067
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,32,8,power_law_1.2,0.11166336059570312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,32,8,power_law_1.2,0.1336678409576416
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,32,8,power_law_1.2,0.16147199630737305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,32,8,power_law_1.2,0.03843712091445923
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,32,8,power_law_1.2,0.07248511791229248
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,32,8,power_law_1.2,0.2225331115722656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,32,8,power_law_1.2,0.025304319858551028
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,32,8,power_law_1.2,0.02624000072479248
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,32,8,power_law_1.2,0.028241920471191406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,32,8,power_law_1.2,0.5224857711791993
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,32,8,power_law_1.2,0.028327679634094237
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,32,8,power_law_1.2,0.029643518924713137
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,32,8,power_law_1.2,0.04787712097167969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,32,8,power_law_1.2,0.04752384185791016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,32,8,power_law_1.2,0.049491200447082516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,32,8,power_law_1.2,0.053647360801696777
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,32,8,power_law_1.2,0.3768076705932617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,32,8,power_law_1.2,0.056266241073608406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,32,8,power_law_1.2,0.05429632186889648
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,32,8,power_law_1.2,0.05149184226989746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,32,8,power_law_1.2,0.057480959892272955
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,32,8,power_law_1.2,0.058880000114440924
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,32,8,power_law_1.2,0.06104576110839843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,32,8,power_law_1.2,0.0673420810699463
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,32,8,power_law_1.2,0.08348928451538086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,32,8,power_law_1.2,0.2678656005859375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,32,8,power_law_1.2,0.1273574447631836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,32,8,power_law_1.2,0.16111488342285157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,32,8,power_law_1.2,0.0512179183959961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,32,8,power_law_1.2,0.9746880340576173
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,32,8,power_law_1.2,0.2674035263061524
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,32,8,power_law_1.2,0.09875840187072754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,32,8,power_law_1.2,0.3947123336791992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,32,8,power_law_1.2,0.21473920822143552
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,8,32,balanced,0.027096319198608398
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,16,16,power_law_1.2,0.9821900939941406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,8,32,balanced,0.02751744031906128
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,8,32,balanced,0.028331520557403567
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,8,32,balanced,0.04967296123504639
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,8,32,balanced,0.09246848106384278
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,8,32,balanced,0.09275391578674316
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,8,32,balanced,0.09622143745422364
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,8,32,balanced,0.09464447975158692
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,8,32,balanced,0.09495807647705078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,8,32,balanced,0.09611647605895997
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,8,32,balanced,0.09788543701171874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,8,32,balanced,0.10066431999206542
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,8,32,balanced,0.10477952003479003
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,8,32,balanced,0.1074944019317627
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,8,32,balanced,0.10992768287658691
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,8,32,balanced,0.10562560081481934
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,8,32,balanced,0.11156479835510254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,8,32,balanced,0.11410176277160644
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,8,32,balanced,0.027388160228729246
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,8,32,balanced,0.1274009609222412
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,8,32,balanced,0.1363814353942871
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,8,32,balanced,0.1594816017150879
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,8,32,balanced,0.17838720321655271
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,8,32,balanced,0.2210099220275879
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,8,32,balanced,0.2672960090637207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,8,32,balanced,0.02790015935897827
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,8,32,balanced,0.36815231323242187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,8,32,balanced,0.028039679527282712
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,8,32,balanced,0.027866880893707275
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,32,8,power_law_1.2,0.9467750549316406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,8,32,balanced,0.028302080631256103
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,8,32,balanced,0.4831897735595703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,8,32,balanced,0.030396161079406736
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,8,32,balanced,0.05252992153167725
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,8,32,balanced,0.05308800220489503
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,8,32,balanced,0.05236480236053467
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,8,32,balanced,0.056916480064392094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,8,32,balanced,0.05589632034301758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,8,32,balanced,0.05679232120513916
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,8,32,balanced,0.05930624008178711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,8,32,balanced,0.06352640151977539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,8,32,balanced,0.06637695789337159
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,32,8,power_law_1.2,0.7631974029541015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,8,32,balanced,0.06919167995452881
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,32,8,power_law_1.2,0.5068262481689454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,8,32,balanced,0.0683289623260498
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,8,32,balanced,0.08022527694702149
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,8,32,balanced,0.09058176040649414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,8,32,balanced,0.11647104263305663
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,8,32,balanced,0.1337830352783203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,8,32,balanced,0.17680511474609376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,8,32,balanced,0.21849599838256836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,8,32,balanced,0.05372032165527344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,8,32,balanced,0.29548032760620113
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,8,32,power_law_1.01,0.052264962196350094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,8,32,balanced,0.37927295684814455
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,8,32,power_law_1.01,0.05260032176971435
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,8,32,power_law_1.01,0.05268864154815674
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,8,32,power_law_1.01,0.05115776062011719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,8,32,balanced,0.06359295845031739
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,8,32,power_law_1.01,0.09063936233520507
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,8,32,power_law_1.01,0.09045632362365723
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,8,32,power_law_1.01,0.09182335853576659
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,8,32,power_law_1.01,0.0982476806640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,8,32,power_law_1.01,0.09783679962158202
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,8,32,power_law_1.01,0.09722240447998047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,8,32,power_law_1.01,0.10268032073974609
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,8,32,power_law_1.01,0.10537856101989747
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,8,32,power_law_1.01,0.10964351654052736
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,8,32,power_law_1.01,0.10901375770568848
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,8,32,power_law_1.01,0.11147775650024414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,8,32,power_law_1.01,0.11040639877319336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,8,32,power_law_1.01,0.12032640457153321
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,8,32,power_law_1.01,0.13330816268920898
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,8,32,power_law_1.01,0.1402956771850586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,8,32,power_law_1.01,0.1624729537963867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,8,32,power_law_1.01,0.19379072189331054
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,8,32,power_law_1.01,0.27810432434082033
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,8,32,power_law_1.01,0.31476224899291994
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,8,32,balanced,0.06965888023376465
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,8,32,power_law_1.01,0.49785728454589845
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,8,32,power_law_1.01,0.03138816118240356
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,8,32,power_law_1.01,0.030943999290466307
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,8,32,power_law_1.01,0.6308544158935547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,8,32,power_law_1.01,0.030023679733276364
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,8,32,power_law_1.01,0.052858881950378425
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,8,32,power_law_1.01,0.03156224012374878
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,32,8,power_law_1.2,0.7471718597412109
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,8,32,power_law_1.01,0.051952638626098634
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,8,32,power_law_1.01,0.05699967861175537
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,8,32,power_law_1.01,0.052980480194091795
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,8,32,power_law_1.01,0.057738242149353025
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,8,32,power_law_1.01,0.06108672142028808
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,8,32,power_law_1.01,0.06403711795806885
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,8,32,power_law_1.01,0.06872320175170898
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,8,32,power_law_1.01,0.065765118598938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,8,32,power_law_1.01,0.06867199897766113
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,8,32,power_law_1.01,0.07570559978485107
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,8,32,power_law_1.01,0.08974847793579102
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,8,32,power_law_1.01,0.10023167610168457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,8,32,power_law_1.01,1.3471653747558594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,8,32,power_law_1.01,0.12051327705383301
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,8,32,power_law_1.01,0.1610598373413086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,8,32,power_law_1.01,0.22775680541992188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,8,32,power_law_1.01,0.2545100784301758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,8,32,power_law_1.01,0.36697982788085937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,8,32,power_law_1.01,0.891115493774414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,8,32,power_law_1.2,0.05246208190917969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,8,32,power_law_1.01,0.5153279876708984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,8,32,power_law_1.01,0.06715263843536376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,8,32,power_law_1.2,0.052259840965271
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,8,32,power_law_1.2,0.05276288032531738
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,8,32,power_law_1.2,0.08737407684326172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,8,32,power_law_1.2,0.09452416419982909
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,8,32,power_law_1.01,0.8215872192382813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,8,32,power_law_1.2,0.09406080245971679
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,8,32,power_law_1.2,0.09669247627258301
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,8,32,power_law_1.01,0.9425933074951172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,8,32,power_law_1.2,0.0998528003692627
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,8,32,power_law_1.2,0.10753664016723632
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,8,32,power_law_1.2,0.10529791831970214
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,8,32,power_law_1.2,0.10967295646667481
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,8,32,power_law_1.2,0.10979455947875975
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,8,32,power_law_1.2,0.10884863853454589
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,8,32,power_law_1.2,0.11662464141845703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,8,32,power_law_1.2,0.12430848121643065
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,8,32,power_law_1.01,0.05844480037689209
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,8,32,power_law_1.2,0.13505663871765136
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,8,32,power_law_1.2,0.15918975830078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,8,32,power_law_1.2,0.19419519424438475
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,8,32,power_law_1.2,0.2170124816894531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,8,32,power_law_1.2,0.08576128005981445
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,8,32,power_law_1.2,0.313176326751709
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,8,32,power_law_1.2,0.09239680290222167
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,8,32,power_law_1.2,0.03158655881881714
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,8,32,power_law_1.2,0.4517926406860352
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,8,32,power_law_1.2,0.030821120738983153
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,8,32,power_law_1.2,0.030227200984954832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,8,32,power_law_1.2,0.03148672103881836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,8,32,power_law_1.2,0.05179647922515869
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,8,32,power_law_1.2,0.05102719783782959
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,8,32,power_law_1.2,0.05453184127807618
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,8,32,power_law_1.2,0.056143360137939455
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,8,32,power_law_1.2,0.05864448070526122
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,8,32,power_law_1.2,0.0605350399017334
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,8,32,power_law_1.2,0.062311677932739264
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,8,32,power_law_1.2,0.06649343967437744
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,8,32,power_law_1.2,0.06804351806640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,8,32,power_law_1.2,0.928305892944336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,8,32,power_law_1.2,0.06931072235107422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,8,32,power_law_1.2,0.07188096046447753
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,8,32,power_law_1.2,0.64801025390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,8,32,power_law_1.2,0.07964928150177002
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,8,32,power_law_1.2,0.09857919692993164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,8,32,power_law_1.2,0.10581888198852538
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,8,32,power_law_1.2,0.1412287998199463
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,8,32,power_law_1.2,0.17349887847900392
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,8,32,power_law_1.2,0.26434047698974605
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,8,32,power_law_1.2,0.3286540985107422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,8,32,power_law_1.2,0.06722559928894042
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,16,16,balanced,0.027060480117797853
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,8,32,power_law_1.2,0.5275660705566406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,16,16,balanced,0.0270361590385437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,16,16,balanced,0.02803071975708008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,16,16,balanced,0.030894079208374024
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,16,16,balanced,0.053116159439086916
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,16,16,balanced,0.09202303886413574
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,8,32,power_law_1.2,0.6856998443603516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,16,16,balanced,0.09456512451171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,16,16,balanced,0.09825407981872558
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,16,16,balanced,0.10050687789916993
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,16,16,balanced,0.10301952362060547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,16,16,balanced,0.10638976097106934
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,16,16,balanced,0.10966400146484374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,16,16,balanced,0.10695679664611817
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,16,16,balanced,0.10837120056152343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,16,16,balanced,0.1125759983062744
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,16,16,balanced,0.1238976001739502
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,16,16,balanced,0.13251968383789064
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,16,16,balanced,0.15293312072753906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,16,16,balanced,0.09629695892333985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,16,16,balanced,0.21097087860107422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,16,16,balanced,0.24954240798950195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,16,16,balanced,0.10363391876220703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,16,16,balanced,0.32323455810546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,16,16,balanced,0.40446720123291013
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,16,16,balanced,0.02803071975708008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,16,16,balanced,0.17097471237182618
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,16,16,balanced,0.028098559379577635
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,16,16,balanced,0.7608844757080078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,16,16,balanced,0.028647680282592774
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,16,16,balanced,0.029885439872741698
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,16,16,balanced,0.03379712104797363
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,16,16,balanced,0.06203648090362549
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,16,16,balanced,0.05903615951538086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,16,16,balanced,0.06032000064849854
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,16,16,balanced,0.06190847873687744
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,16,16,balanced,0.06303359985351562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,16,16,balanced,0.06711167812347411
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,16,16,balanced,0.06852096080780029
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,16,16,balanced,0.07298687934875489
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,8,32,power_law_1.2,1.0280754852294922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,16,16,balanced,0.06725632190704346
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,16,16,balanced,0.06884352207183839
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,8,32,power_law_1.2,1.3426124572753906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,16,16,balanced,0.08253439903259277
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,16,16,balanced,0.09296256065368652
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,16,16,balanced,0.11201279640197755
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,16,16,balanced,0.12994303703308105
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,16,16,balanced,0.17529855728149416
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,16,16,balanced,0.21087488174438476
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,16,16,balanced,0.2896678352355957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,16,16,balanced,0.3699276733398437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,16,16,balanced,0.5800332641601562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,16,16,power_law_1.01,0.05102591991424561
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,16,16,balanced,0.07314688205718994
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,16,16,power_law_1.01,0.05353343963623047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,16,16,power_law_1.01,0.05508607864379883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,16,16,balanced,0.7079283142089844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,16,16,power_law_1.01,0.055776000022888184
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,8,32,power_law_1.2,1.364515838623047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,16,16,power_law_1.01,0.08941823959350585
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,16,16,power_law_1.01,0.09120896339416504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,16,16,power_law_1.01,0.09395839691162108
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,16,16,power_law_1.01,0.0968064022064209
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,16,16,power_law_1.01,0.10072064399719238
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,16,16,power_law_1.01,0.10543231964111328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,16,16,power_law_1.01,0.09987071990966798
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,16,16,power_law_1.01,0.10902655601501465
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,16,16,power_law_1.01,0.11165311813354492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,16,16,power_law_1.01,0.10621824264526367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,16,16,power_law_1.01,0.11066623687744141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,16,16,power_law_1.01,0.11361663818359376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,16,16,power_law_1.01,0.1224998378753662
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,16,16,power_law_1.01,0.13760640144348144
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,16,16,power_law_1.01,0.15492735862731935
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,16,16,power_law_1.01,0.1761075210571289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,16,16,power_law_1.01,0.1963737678527832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,16,16,power_law_1.01,0.2632947158813477
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,16,16,power_law_1.01,0.30836992263793944
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,16,16,power_law_1.01,0.4783283233642578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,16,16,power_law_1.01,0.6471360015869141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,16,16,power_law_1.01,0.0321612811088562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,16,16,balanced,0.5183027267456055
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,16,16,power_law_1.01,0.03502336025238037
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,16,16,power_law_1.01,0.03403264045715332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,16,16,power_law_1.01,0.9961689758300782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,16,16,power_law_1.01,0.03652607917785645
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,16,16,power_law_1.01,0.054004478454589847
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,16,16,power_law_1.01,0.060995841026306154
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,16,16,power_law_1.01,0.059781122207641604
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,16,16,power_law_1.01,0.06292479991912842
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,16,16,power_law_1.01,0.06566400051116943
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,16,16,power_law_1.01,0.0680076789855957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,16,16,power_law_1.01,0.0736191987991333
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,16,16,power_law_1.01,0.07149951934814454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,16,16,power_law_1.01,0.06968832015991211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,16,16,balanced,0.07088255882263184
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,16,16,power_law_1.01,0.07230591773986816
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,16,16,power_law_1.01,0.07375487804412842
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,16,16,power_law_1.01,0.08354304313659669
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,16,16,power_law_1.01,0.09753087997436524
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,16,16,power_law_1.01,0.10637824058532715
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,16,16,power_law_1.01,0.14283007621765137
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,16,16,power_law_1.01,0.05989503860473633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,16,16,power_law_1.01,0.1814259147644043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,16,16,power_law_1.01,0.22871551513671876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,16,16,power_law_1.01,1.279073257446289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,16,16,power_law_1.01,0.3107289505004883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,8,32,power_law_1.2,1.9991552734374998
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,16,16,power_law_1.2,0.05158527851104736
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,16,16,power_law_1.01,0.42249599456787107
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,16,16,power_law_1.2,0.056337919235229496
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,16,16,power_law_1.2,0.05316351890563965
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,16,16,power_law_1.01,0.4943654251098633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,16,16,power_law_1.2,0.05515520095825195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,16,16,power_law_1.2,0.0877120018005371
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,16,16,power_law_1.2,0.09076736450195313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,16,16,power_law_1.2,0.10194175720214844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,16,16,power_law_1.2,0.10772224426269532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,16,16,power_law_1.01,0.8663616180419922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,16,16,power_law_1.2,0.11085311889648437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,16,16,power_law_1.2,0.10693120002746581
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,16,16,power_law_1.2,0.1089907169342041
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,16,16,power_law_1.2,0.1111308765411377
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,16,16,power_law_1.2,0.11609984397888184
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,16,16,power_law_1.2,0.12568448066711427
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,16,16,power_law_1.2,0.14330368041992186
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,16,16,power_law_1.2,0.16092544555664062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,16,16,power_law_1.2,0.19186304092407228
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,16,16,power_law_1.2,0.23194623947143556
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,16,16,power_law_1.2,0.30810239791870114
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,16,16,power_law_1.2,0.09477248191833496
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,16,16,power_law_1.2,0.09852800369262696
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,16,16,power_law_1.2,0.3886912155151367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,16,16,power_law_1.2,0.032257280349731444
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,16,16,power_law_1.01,1.0946892547607423
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,16,16,power_law_1.2,0.5717875289916992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,16,16,power_law_1.2,0.034082560539245604
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,16,16,power_law_1.2,0.03578880071640015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,16,16,power_law_1.2,0.03544960021972656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,16,16,power_law_1.2,0.053497600555419925
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,16,16,power_law_1.2,0.05748991966247559
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,16,16,power_law_1.2,0.0638963222503662
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,16,16,power_law_1.2,0.060359678268432616
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,16,16,power_law_1.2,0.06714879989624023
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,16,16,power_law_1.2,0.06955520153045655
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,16,16,power_law_1.2,0.09492608070373534
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,16,16,power_law_1.2,0.07197184085845947
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,16,16,power_law_1.2,0.06884736061096192
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,16,16,power_law_1.2,0.07058815956115723
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,16,16,power_law_1.2,0.07421951770782471
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,16,16,power_law_1.2,0.07570816040039062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,16,16,power_law_1.2,0.08409728050231934
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,16,16,power_law_1.2,0.0987558364868164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,16,16,power_law_1.2,0.11358336448669433
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,16,16,power_law_1.2,0.16868480682373047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,16,16,power_law_1.2,0.18202239990234376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,16,16,power_law_1.2,0.25147520065307616
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,16,16,power_law_1.2,0.06265088081359863
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,16,16,power_law_1.2,0.3265267181396484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,32,8,balanced,0.027444479465484617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,32,8,balanced,0.02853503942489624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,16,16,power_law_1.2,0.5837324905395508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,16,16,power_law_1.2,0.5031744003295898
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,32,8,balanced,0.030608639717102048
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,32,8,balanced,0.041196799278259276
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,32,8,balanced,0.05822720050811767
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,32,8,balanced,0.09983872413635254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,32,8,balanced,0.10257920265197754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,32,8,balanced,0.10655232429504395
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,32,8,balanced,0.10836223602294921
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,32,8,balanced,0.10523391723632813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,16,16,power_law_1.2,0.7959705352783203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,32,8,balanced,0.10809087753295898
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,32,8,balanced,0.11018624305725097
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,32,8,balanced,0.11859968185424805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,32,8,balanced,0.11503616333007813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,32,8,balanced,0.1242739200592041
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,32,8,balanced,0.13293824195861817
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,32,8,balanced,0.1508556842803955
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,32,8,balanced,0.16832000732421876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,32,8,balanced,0.21020927429199218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,32,8,balanced,0.10908543586730957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,32,8,balanced,0.24451072692871093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,32,8,balanced,0.32037376403808593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,32,8,balanced,0.3956467056274414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,16,16,power_law_1.2,1.158054428100586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,32,8,balanced,0.5477798461914063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,32,8,balanced,0.028247039318084716
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,32,8,balanced,0.7034175872802735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,32,8,balanced,0.02895103931427002
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,32,8,balanced,0.030018560886383057
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,32,8,balanced,0.03371392011642456
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,32,8,balanced,0.03968767881393433
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,32,8,balanced,0.06935296058654786
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,32,8,balanced,1.0553778839111327
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,32,8,balanced,0.06966911792755127
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,32,8,balanced,0.0737779188156128
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,32,8,balanced,0.07918975830078126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,32,8,balanced,0.07659776210784912
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,32,8,balanced,0.07297023773193359
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,32,8,balanced,0.07602303981781006
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,32,8,balanced,0.07830016136169433
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,32,8,balanced,1.3883686828613282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,32,8,balanced,0.08288000106811524
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,32,8,balanced,0.08423680305480957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,32,8,balanced,0.08850432395935058
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,32,8,balanced,0.0982271957397461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,32,8,balanced,0.11765376091003418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,32,8,balanced,0.13559167861938476
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,32,8,balanced,0.17134592056274414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,32,8,balanced,0.2075200080871582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,32,8,balanced,0.29029375076293945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,32,8,balanced,0.35961471557617186
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,32,8,balanced,0.5121779251098633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,32,8,power_law_1.01,0.031371519565582276
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,32,8,balanced,0.6653158569335937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,32,8,power_law_1.01,0.053706240653991696
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,32,8,power_law_1.01,0.05176959991455078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,16,16,power_law_1.2,1.7080192565917969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,16,16,power_law_1.2,1.3362777709960938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,32,8,balanced,0.9564582061767577
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,32,8,power_law_1.01,0.057556481361389154
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,32,8,power_law_1.01,0.09676671981811523
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,32,8,power_law_1.01,0.09841792106628418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,32,8,power_law_1.01,0.10264960289001465
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,32,8,power_law_1.01,0.10595968246459961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,32,8,power_law_1.01,0.11221504211425781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,32,8,balanced,1.2956863403320313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,32,8,power_law_1.01,0.11219712257385255
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,32,8,power_law_1.01,0.10705663681030272
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,32,8,power_law_1.01,0.11098496437072754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,32,8,power_law_1.01,0.11342592239379883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,32,8,power_law_1.01,0.11741696357727052
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,32,8,power_law_1.01,0.14368895530700682
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,32,8,power_law_1.01,0.16040447235107422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,32,8,power_law_1.01,0.19574016571044922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,32,8,power_law_1.01,0.05566207885742187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,32,8,power_law_1.01,0.2263987159729004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,32,8,power_law_1.01,0.2869964790344238
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,32,8,power_law_1.01,0.3702643203735351
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,32,8,power_law_1.01,0.1068671989440918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,32,8,power_law_1.01,0.03224191904067993
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,32,8,power_law_1.01,0.03630208015441895
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,32,8,power_law_1.01,0.03529983997344971
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,32,8,power_law_1.01,0.12615039825439453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,32,8,power_law_1.01,0.03912575960159302
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,32,8,power_law_1.01,0.04226816177368164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,32,8,power_law_1.01,0.06479872226715087
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,32,8,power_law_1.01,1.0668287658691407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,32,8,power_law_1.01,0.06970880031585694
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,32,8,power_law_1.01,0.07400959968566895
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,32,8,power_law_1.01,0.07485951900482177
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,32,8,power_law_1.01,1.3401075744628907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,32,8,power_law_1.01,0.07946752071380615
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,32,8,power_law_1.01,0.07495552062988281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,32,8,power_law_1.01,0.5424832153320313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,32,8,power_law_1.01,0.07392384052276611
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,32,8,power_law_1.01,0.07986688137054443
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,32,8,power_law_1.01,0.0785472011566162
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,32,8,power_law_1.01,0.08256640434265136
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,32,8,power_law_1.01,0.6851699066162109
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,32,8,power_law_1.01,0.11209856033325197
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,32,8,power_law_1.01,0.13436415672302246
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,32,8,power_law_1.01,0.1733951950073242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,32,8,power_law_1.01,0.2031692886352539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,32,8,power_law_1.01,0.06541312217712403
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,32,8,power_law_1.01,0.25696128845214844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,32,8,power_law_1.01,0.3528704071044922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,32,8,power_law_1.2,0.03105407953262329
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,32,8,power_law_1.01,0.4963827133178711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,32,8,power_law_1.01,0.5671142578125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,32,8,power_law_1.2,0.05238656044006348
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,32,8,power_law_1.01,0.09089792251586915
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,32,8,power_law_1.2,0.05578752040863037
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,32,8,power_law_1.2,0.05722879886627197
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,32,8,power_law_1.2,0.09579263687133789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,32,8,power_law_1.01,0.9964864349365234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,32,8,power_law_1.2,0.09928192138671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,32,8,power_law_1.2,0.1045081615447998
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,32,8,power_law_1.2,0.10891263961791993
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,32,8,power_law_1.2,0.11215999603271484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,32,8,power_law_1.2,0.10946944236755371
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,32,8,power_law_1.2,0.1080038356781006
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,32,8,power_law_1.01,1.3748556518554689
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,32,8,power_law_1.2,0.11207039833068848
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,32,8,power_law_1.2,0.11422335624694824
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,32,8,power_law_1.2,0.049523200988769535
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,32,8,power_law_1.2,0.12132351875305177
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,32,8,power_law_1.2,0.1347315216064453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,32,8,power_law_1.2,0.14672767639160156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,32,8,power_law_1.2,0.1665216064453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,32,8,power_law_1.2,0.2090982437133789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,32,8,power_law_1.2,0.24664064407348635
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,32,8,power_law_1.2,0.10153471946716308
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,32,8,power_law_1.2,0.3090687942504883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,32,8,power_law_1.2,0.39789054870605467
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,32,8,power_law_1.2,0.032084479331970214
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,32,8,power_law_1.2,0.03365504026412964
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,32,8,power_law_1.2,0.5691238403320312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,32,8,power_law_1.2,0.04030335903167724
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,32,8,power_law_1.2,0.038526721000671386
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,32,8,power_law_1.2,0.06775680065155029
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,16,16,power_law_1.2,0.9580185699462891
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,32,8,power_law_1.2,0.07038080215454101
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,32,8,power_law_1.2,0.07289472103118896
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,32,8,power_law_1.2,0.07414400100708007
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,32,8,power_law_1.2,0.03512703895568848
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,32,8,power_law_1.2,0.07902336120605469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,32,8,power_law_1.2,0.7958271789550782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,32,8,power_law_1.2,0.07502336025238036
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,32,8,power_law_1.2,0.07469312191009522
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,32,8,power_law_1.2,0.0796236801147461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,32,8,power_law_1.2,0.0816140842437744
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,32,8,power_law_1.2,0.09324799537658691
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,32,8,power_law_1.2,0.08644607543945312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,32,8,power_law_1.2,1.1469042968749998
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,32,8,power_law_1.2,0.11541888236999512
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,32,8,power_law_1.2,0.13420160293579103
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,32,8,power_law_1.2,0.17522943496704102
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,32,8,power_law_1.2,0.06568831920623779
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,32,8,power_law_1.2,0.21536256790161135
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,32,8,power_law_1.2,0.2748415946960449
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,2,128,balanced,0.056076798439025875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,32,8,power_law_1.2,0.3617049789428711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,2,128,balanced,0.056055040359497064
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,2,128,balanced,0.05597184181213379
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,2,128,balanced,0.05078911781311035
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,2,128,balanced,0.05561600208282471
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,32,8,power_law_1.2,0.46480640411376956
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,2,128,balanced,0.06883584022521973
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,2,128,balanced,0.11864319801330567
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,2,128,balanced,0.11926527976989747
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,2,128,balanced,0.11922816276550292
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,2,128,balanced,0.12050944328308107
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,32,8,power_law_1.2,0.9751602935791016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,2,128,balanced,0.12229632377624511
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,2,128,balanced,0.12228608131408691
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,2,128,balanced,0.12674176216125488
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,2,128,balanced,0.1244275188446045
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,2,128,balanced,0.12582528114318847
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,2,128,balanced,0.1311411190032959
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,2,128,balanced,0.13494272232055665
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,2,128,balanced,0.141593599319458
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,2,128,balanced,0.13939071655273438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,2,128,balanced,0.06296576023101806
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,2,128,balanced,0.14640512466430664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,2,128,balanced,0.15550463676452636
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,2,128,balanced,0.16938880920410154
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,2,128,balanced,0.18549888610839843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,2,128,balanced,0.22599679946899415
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,2,128,balanced,0.03164160013198852
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,2,128,balanced,0.0317196798324585
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,2,128,balanced,0.3054707145690918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,32,8,power_law_1.2,0.6828876495361328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,2,128,balanced,0.03184384107589722
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,2,128,balanced,0.0317248010635376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,2,128,balanced,0.03172863960266113
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,2,128,balanced,0.03821439981460571
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,2,128,balanced,0.044398078918457026
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,2,128,balanced,0.06084735870361328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,2,128,balanced,0.061236481666564944
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,2,128,balanced,0.06021376132965088
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,2,128,balanced,0.061354241371154784
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,2,128,balanced,0.06140416145324708
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,2,128,balanced,0.06250368118286133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,2,128,balanced,0.066494722366333
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,2,128,balanced,0.06396671772003173
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,2,128,balanced,0.06462592124938965
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,2,128,balanced,0.06859903812408448
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,2,128,balanced,0.07147264003753662
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,2,128,balanced,0.0776473617553711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,2,128,balanced,0.0729804801940918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,2,128,balanced,0.07967743873596192
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,2,128,balanced,0.08401663780212403
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,2,128,balanced,0.09735936164855957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,2,128,balanced,0.11330304145812989
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,2,128,balanced,0.12165120124816894
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,2,128,balanced,0.16771968841552734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,2,128,power_law_1.01,0.06931327819824219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,2,128,power_law_1.01,0.06873216152191161
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,2,128,power_law_1.01,0.06985856056213378
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,2,128,power_law_1.01,0.069203200340271
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,2,128,power_law_1.01,0.1181824016571045
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,2,128,balanced,0.0664844799041748
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,2,128,power_law_1.01,0.11577088356018066
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,2,128,power_law_1.01,0.1180940818786621
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,32,8,power_law_1.2,1.3343820190429687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,2,128,power_law_1.01,0.12133888244628906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,2,128,power_law_1.01,0.12223103523254394
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,2,128,balanced,0.13847167968750002
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,2,128,power_law_1.01,0.1294553565979004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,2,128,power_law_1.01,0.12514047622680663
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,2,128,power_law_1.01,0.13058815956115724
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,2,128,power_law_1.01,0.13516544342041015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,2,128,power_law_1.01,0.1370483207702637
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,2,128,power_law_1.01,0.14125311851501465
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,2,128,power_law_1.01,0.14763520240783692
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,2,128,power_law_1.01,0.15489919662475587
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,2,128,power_law_1.01,0.17440000534057618
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,2,128,power_law_1.01,0.25997951507568356
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,2,128,power_law_1.01,0.3411507034301758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,2,128,power_law_1.01,0.12408320426940918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,2,128,power_law_1.01,0.48671745300292973
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,2,128,power_law_1.01,0.5602790451049804
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,2,128,power_law_1.01,0.044015359878540036
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,2,128,power_law_1.01,0.04452735900878906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,2,128,power_law_1.01,0.2067750358581543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,2,128,power_law_1.01,0.044247040748596186
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,2,128,power_law_1.01,0.9672614288330077
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,2,128,power_law_1.01,0.058880000114440924
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,2,128,power_law_1.01,0.12101887702941894
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,2,128,power_law_1.01,0.05922944068908691
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,2,128,power_law_1.01,1.1726873779296876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,2,128,power_law_1.01,0.06096511840820312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,2,128,power_law_1.01,0.0625113582611084
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,2,128,power_law_1.01,0.06184447765350341
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,2,128,power_law_1.01,0.06268288135528564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,2,128,power_law_1.01,0.06458752155303955
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,2,128,power_law_1.01,0.06632063865661622
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,2,128,power_law_1.01,0.06963712215423584
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,2,128,power_law_1.01,0.06962175846099854
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,2,128,power_law_1.01,0.07270016193389892
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,2,128,power_law_1.01,0.07532288074493408
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,2,128,power_law_1.01,0.0800755214691162
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,2,128,power_law_1.01,0.0858675193786621
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,2,128,power_law_1.01,0.10099712371826172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,2,128,power_law_1.01,0.11410176277160644
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,2,128,power_law_1.01,0.044465918540954594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,2,128,power_law_1.01,0.13330304145812988
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,2,128,power_law_1.01,0.18203647613525392
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,2,128,power_law_1.01,0.2478144073486328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,2,128,power_law_1.2,0.0688755178451538
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,2,128,power_law_1.2,0.06895359992980957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,2,128,power_law_1.01,0.3560755157470703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,2,128,power_law_1.2,0.06980095863342285
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,2,128,power_law_1.01,0.5015398406982422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,2,128,power_law_1.2,0.06936960220336914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,2,128,power_law_1.2,0.11450112342834473
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,2,128,power_law_1.2,0.11648639678955078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,2,128,power_law_1.01,0.0796454381942749
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,2,128,power_law_1.2,0.11801343917846681
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,2,128,power_law_1.2,0.11920512199401856
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,2,128,power_law_1.2,0.11925503730773926
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,2,128,power_law_1.2,0.11847807884216308
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,2,128,power_law_1.2,0.12182527542114259
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,2,128,power_law_1.2,0.12544511795043944
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,2,128,power_law_1.2,0.1354188823699951
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,2,128,power_law_1.2,0.1278604793548584
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,2,128,power_law_1.2,0.13778047561645507
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,2,128,power_law_1.2,0.1413913631439209
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,2,128,power_law_1.2,0.1588326358795166
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,2,128,power_law_1.2,0.17257984161376955
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,2,128,power_law_1.2,0.20593664169311526
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,2,128,power_law_1.2,0.2575974464416504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,2,128,power_law_1.2,0.3510041427612305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,2,128,power_law_1.01,0.6651302337646484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,2,128,power_law_1.2,0.4741107177734375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,2,128,power_law_1.2,0.0444979190826416
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,2,128,power_law_1.2,0.12317695617675781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,2,128,power_law_1.2,0.6776217651367188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,2,128,power_law_1.2,0.044550399780273434
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,2,128,power_law_1.2,0.04425600051879883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,2,128,power_law_1.2,0.9456333160400391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,2,128,power_law_1.2,0.04455552101135254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,2,128,power_law_1.2,0.05886847972869873
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,2,128,power_law_1.2,0.06177408218383788
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,2,128,power_law_1.2,0.061194238662719724
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,2,128,power_law_1.2,0.060971522331237794
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,2,128,power_law_1.2,0.06323584079742431
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,2,128,power_law_1.2,0.07012864112854003
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,2,128,power_law_1.2,1.3853503417968749
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,2,128,power_law_1.2,0.06610943794250487
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,2,128,power_law_1.2,0.06673408031463623
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,2,128,power_law_1.2,0.06969600200653077
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,2,128,power_law_1.2,0.07322495937347412
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,2,128,power_law_1.2,0.07479680061340332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,2,128,power_law_1.2,0.07800831794738769
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,2,128,power_law_1.2,0.09042176246643066
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,2,128,power_law_1.2,0.09594367980957032
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,2,128,power_law_1.2,0.1254310417175293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,2,128,power_law_1.2,0.1329792022705078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,32,8,power_law_1.2,1.5635264587402342
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,2,128,power_law_1.2,1.9850367736816406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,2,128,power_law_1.2,0.1900876808166504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,2,128,power_law_1.2,0.23870080947875977
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,4,64,balanced,0.030696959495544435
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,2,128,power_law_1.2,0.36267265319824216
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,4,64,balanced,0.030645759105682374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,4,64,balanced,0.030720000267028806
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,2,128,power_law_1.2,0.4986483383178711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,4,64,balanced,0.03069567918777466
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,4,64,balanced,0.053822717666625976
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,4,64,balanced,0.06162816047668457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,4,64,balanced,0.06958591938018799
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,4,64,balanced,0.11634943962097169
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,4,64,balanced,0.11392512321472167
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,4,64,balanced,0.11415167808532715
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,4,64,balanced,0.1161689567565918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,4,64,balanced,0.12421119689941407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,4,64,balanced,0.118984956741333
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,2,128,power_law_1.2,0.9435865783691406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,4,64,balanced,0.12234111785888672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,2,128,power_law_1.2,1.0356326293945313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,4,64,balanced,0.12455552101135255
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,4,64,balanced,0.12579456329345703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,4,64,balanced,0.12857215881347656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,4,64,balanced,0.13247488021850587
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,4,64,balanced,0.13468288421630858
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,4,64,balanced,0.14750080108642577
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,4,64,balanced,0.15917440414428713
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,4,64,balanced,0.17611391067504883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,2,128,power_law_1.2,0.061976318359375004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,4,64,balanced,0.1963737678527832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,4,64,balanced,0.22303487777709963
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,4,64,balanced,0.03058432102203369
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,4,64,balanced,0.27567487716674804
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,4,64,balanced,0.030589439868927003
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,4,64,balanced,0.03053440093994141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,4,64,balanced,0.37566207885742187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,4,64,balanced,0.031111679077148437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,4,64,balanced,0.03892224073410035
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,4,64,balanced,0.06515711784362793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,4,64,balanced,0.06460415840148925
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,4,64,balanced,0.066975998878479
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,4,64,balanced,0.06796927928924561
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,4,64,balanced,0.06654592037200928
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,4,64,balanced,0.06808063983917237
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,4,64,balanced,0.06935935974121094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,4,64,balanced,0.07014912128448486
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,4,64,balanced,0.0736844778060913
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,4,64,balanced,0.14031871795654297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,4,64,balanced,0.07802752017974854
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,4,64,balanced,0.07306367874145507
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,4,64,balanced,0.07813375949859619
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,4,64,balanced,0.08139007568359376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,4,64,balanced,0.09336319923400879
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,4,64,balanced,0.10651391983032225
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,4,64,balanced,0.03074559926986694
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,4,64,balanced,0.12956031799316406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,4,64,balanced,0.15875840187072754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,4,64,balanced,0.20024831771850585
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,4,64,power_law_1.01,0.06839551925659179
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,4,64,balanced,0.2514329528808594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,4,64,power_law_1.01,0.06810111999511718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,4,64,power_law_1.01,0.06895999908447266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,4,64,balanced,0.06751488208770752
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,4,64,power_law_1.01,0.11145983695983888
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,4,64,power_law_1.01,0.11062272071838379
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,4,64,power_law_1.01,0.11390848159790039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,4,64,power_law_1.01,0.11561216354370116
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,4,64,power_law_1.01,0.12295552253723145
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,4,64,power_law_1.01,0.11854592323303223
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,4,64,power_law_1.01,0.12242048263549805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,4,64,power_law_1.01,0.12402815818786621
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,4,64,power_law_1.01,0.13277055740356444
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,4,64,power_law_1.01,0.13557375907897948
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,4,64,power_law_1.01,0.14743167877197266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,4,64,power_law_1.01,0.15661312103271485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,4,64,balanced,0.047470078468322754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,4,64,power_law_1.01,0.17322751998901367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,4,64,power_law_1.01,0.06968832015991211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,4,64,power_law_1.01,0.20200191497802736
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,4,64,power_law_1.01,0.2692927932739258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,4,64,power_law_1.01,0.11401599884033202
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,4,64,power_law_1.01,0.3279206466674805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,4,64,power_law_1.01,0.39114753723144535
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,4,64,power_law_1.01,0.04686336040496826
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,4,64,power_law_1.01,0.12827775955200196
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,4,64,power_law_1.01,0.04623871803283691
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,4,64,power_law_1.01,0.13856512069702148
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,4,64,power_law_1.01,0.04656896114349365
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,4,64,power_law_1.01,0.047320318222045896
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,4,64,power_law_1.01,0.06397183895111083
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,4,64,power_law_1.01,0.06346879959106445
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,4,64,power_law_1.01,0.061918721199035645
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,4,64,power_law_1.01,0.06697472095489501
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,4,64,power_law_1.01,0.8586444854736328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,4,64,power_law_1.01,0.06738560199737549
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,4,64,power_law_1.01,0.06742911815643311
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,4,64,power_law_1.01,0.06960768222808837
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,4,64,power_law_1.01,0.07040512084960937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,4,64,power_law_1.01,0.07481472015380859
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,4,64,power_law_1.01,0.07941120147705079
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,4,64,power_law_1.01,0.08190079689025878
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,4,64,power_law_1.01,0.537918701171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,4,64,power_law_1.01,0.0876966381072998
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,4,64,power_law_1.01,0.10130816459655763
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,4,64,power_law_1.01,0.13041791915893555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,4,64,power_law_1.01,0.832147216796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,4,64,power_law_1.01,0.16104959487915038
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,4,64,power_law_1.01,0.20877439498901368
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,4,64,power_law_1.01,0.06750592231750488
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,4,64,power_law_1.01,0.26071935653686523
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,4,64,power_law_1.01,0.33401729583740236
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,4,64,power_law_1.2,0.06860032081604003
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,4,64,power_law_1.01,0.07686655998229981
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,4,64,power_law_1.01,0.4883801651000977
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,4,64,power_law_1.01,0.08412672042846679
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,4,64,power_law_1.2,0.0693235206604004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,4,64,power_law_1.01,0.6190028762817382
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,4,64,power_law_1.2,0.11242752075195313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,4,64,power_law_1.2,0.10989312171936035
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,4,64,power_law_1.2,0.11224575996398926
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,4,64,power_law_1.2,0.11514495849609374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,4,64,power_law_1.2,0.11789055824279786
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,4,64,power_law_1.2,0.11725184440612793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,4,64,power_law_1.2,0.12408320426940918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,4,64,power_law_1.2,0.12564736366271972
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,4,64,power_law_1.2,0.13577856063842772
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,4,64,power_law_1.2,0.13297151565551757
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,4,64,power_law_1.2,0.06805759906768799
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,4,64,power_law_1.2,0.14010239601135255
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,4,64,power_law_1.2,0.14637184143066406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,4,64,power_law_1.2,0.0697215986251831
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,4,64,power_law_1.2,0.16592767715454101
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,4,64,power_law_1.2,0.17992448806762695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,4,64,power_law_1.2,0.22421503067016602
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,4,64,power_law_1.2,0.2994764709472656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,4,64,power_law_1.2,0.38266624450683595
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,4,64,power_law_1.2,0.11907839775085449
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,4,64,power_law_1.2,0.04665215969085693
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,4,64,power_law_1.2,0.04644480228424072
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,4,64,power_law_1.2,0.6812493133544921
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,4,64,power_law_1.2,0.13046400070190428
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,4,64,power_law_1.2,0.04691840171813964
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,4,64,power_law_1.2,0.048060159683227535
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,4,64,power_law_1.2,0.7998566436767578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,4,64,power_law_1.2,0.06371967792510987
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,4,64,power_law_1.2,0.06265600204467774
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,4,64,power_law_1.2,0.0653107213973999
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,4,64,power_law_1.2,1.1136946868896485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,4,64,power_law_1.2,0.06785408020019532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,4,64,power_law_1.2,0.07380095958709718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,4,64,power_law_1.2,0.06807936191558837
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,4,64,power_law_1.2,0.0686195182800293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,4,64,power_law_1.2,0.07091455936431885
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,4,64,power_law_1.2,0.07714431762695313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,4,64,power_law_1.2,0.07518335819244384
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,4,64,power_law_1.2,0.07509759902954102
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,4,64,power_law_1.2,1.4784640502929687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,4,64,power_law_1.2,0.07921664237976074
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,4,64,power_law_1.2,0.08484992027282715
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,4,64,power_law_1.2,0.10346240043640136
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,4,64,power_law_1.2,0.11858304023742676
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,4,64,power_law_1.2,0.15611647605895995
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,4,64,power_law_1.2,0.17777280807495116
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,4,64,power_law_1.2,0.24385152816772462
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,8,32,balanced,0.030304000377655026
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,8,32,balanced,0.029931519031524655
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,8,32,balanced,0.030196480751037598
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,8,32,balanced,0.03142911911010742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,4,64,power_law_1.2,0.060579838752746584
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,8,32,balanced,0.05495423793792724
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,4,64,power_law_1.2,0.5206655883789063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,8,32,balanced,0.06388735771179199
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,8,32,balanced,0.07197824001312256
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,8,32,balanced,0.1198796844482422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,8,32,balanced,0.12468480110168456
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,8,32,balanced,0.12302207946777344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,8,32,balanced,0.12653311729431152
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,8,32,balanced,0.11971072196960449
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,8,32,balanced,0.1265510368347168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,8,32,balanced,0.13266048431396485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,8,32,balanced,0.13380736351013184
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,8,32,balanced,0.13713536262512208
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,4,64,power_law_1.2,0.9728050994873048
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,8,32,balanced,0.1376793575286865
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,8,32,balanced,0.14335488319396972
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,8,32,balanced,0.1544102382659912
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,8,32,balanced,0.16501632690429688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,8,32,balanced,0.18672128677368166
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,8,32,balanced,0.25224191665649415
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,4,64,power_law_1.2,0.37870079040527344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,8,32,balanced,0.2990336036682129
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,8,32,balanced,0.03009151935577393
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,8,32,balanced,0.39988609313964846
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,8,32,balanced,0.030314240455627438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,8,32,balanced,0.031020801067352295
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,8,32,balanced,0.03278464078903198
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,8,32,balanced,0.5276646423339844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,4,64,power_law_1.2,0.824865264892578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,8,32,balanced,0.04426496028900147
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,8,32,balanced,0.05078144073486328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,8,32,balanced,0.0730521583557129
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,8,32,balanced,0.07585408210754394
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,8,32,balanced,0.07155839920043945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,8,32,balanced,0.07190783977508544
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,8,32,balanced,0.07460991859436035
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,8,32,balanced,0.0732480001449585
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,8,32,balanced,0.07918335914611815
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,8,32,balanced,0.21082496643066406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,8,32,balanced,0.079966721534729
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,8,32,balanced,0.08240384101867676
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,8,32,balanced,0.0819865608215332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,8,32,balanced,0.08546688079833983
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,8,32,balanced,0.09632512092590331
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,8,32,balanced,0.10822655677795409
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,8,32,balanced,0.12855680465698244
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,8,32,balanced,0.03015552043914795
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,8,32,balanced,0.15299455642700194
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,8,32,balanced,0.19128576278686524
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,8,32,balanced,0.24501504898071289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,8,32,power_law_1.01,0.06345088005065917
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,8,32,balanced,0.1311180782318115
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,8,32,balanced,0.324766731262207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,8,32,power_law_1.01,0.06967423915863037
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,8,32,balanced,0.4176102447509765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,8,32,power_law_1.01,0.0702681589126587
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,8,32,power_law_1.01,0.11672191619873047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,8,32,power_law_1.01,0.11456640243530274
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,8,32,balanced,0.0776255989074707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,8,32,power_law_1.01,0.11882240295410156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,8,32,power_law_1.01,0.11814528465270997
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,8,32,power_law_1.01,0.12221055984497071
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,8,32,power_law_1.01,0.1294003200531006
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,8,32,power_law_1.01,0.12783616065979003
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,8,32,power_law_1.01,0.1333606433868408
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,8,32,power_law_1.01,0.13144703865051272
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,8,32,power_law_1.01,0.13110655784606934
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,8,32,power_law_1.01,0.13784704208374024
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,8,32,power_law_1.01,0.15100928306579592
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,8,32,power_law_1.01,0.15261055946350097
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,8,32,power_law_1.01,0.0705728006362915
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,8,32,power_law_1.01,0.20135040283203126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,8,32,power_law_1.01,0.11258111953735353
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,8,32,power_law_1.01,0.2591168022155762
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,8,32,power_law_1.01,0.3035123252868652
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,8,32,power_law_1.01,0.11642623901367188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,8,32,power_law_1.01,0.4462847900390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,8,32,power_law_1.01,0.04004608154296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,8,32,power_law_1.01,0.49521152496337895
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,8,32,power_law_1.01,0.04838655948638916
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,8,32,power_law_1.01,0.04852863788604737
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,8,32,power_law_1.01,0.05077760219573975
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,8,32,power_law_1.01,0.7449100494384766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,8,32,power_law_1.01,0.06993663787841797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,8,32,power_law_1.01,0.18199424743652343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,8,32,power_law_1.01,0.0708787202835083
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,8,32,power_law_1.01,0.07212416172027589
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,8,32,power_law_1.01,1.0033715057373047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,8,32,power_law_1.01,0.07459199905395508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,8,32,power_law_1.01,0.07050623893737792
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,8,32,power_law_1.01,0.07780992031097413
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,8,32,power_law_1.01,0.08319871902465821
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,8,32,power_law_1.01,0.07883647918701173
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,8,32,power_law_1.01,0.08355456352233886
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,8,32,power_law_1.01,0.07927296161651612
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,8,32,power_law_1.01,0.09272576332092285
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,8,32,power_law_1.01,0.09875328063964843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,8,32,power_law_1.01,0.12333951950073244
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,8,32,power_law_1.01,0.1430617618560791
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,8,32,power_law_1.01,0.04961408138275146
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,8,32,power_law_1.01,0.19650047302246093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,8,32,power_law_1.01,0.24025344848632812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,8,32,power_law_1.01,0.3106931114196777
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,8,32,power_law_1.2,0.06321792125701904
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,8,32,power_law_1.2,0.06951807975769043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,8,32,power_law_1.01,0.4431615829467773
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,8,32,power_law_1.01,0.07485439777374267
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,8,32,power_law_1.01,0.6134592056274414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,8,32,power_law_1.2,0.07179008007049562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,8,32,power_law_1.2,0.1103065586090088
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,8,32,power_law_1.2,0.11661696434020996
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,8,32,power_law_1.2,0.11524736404418945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,8,32,power_law_1.2,0.11953920364379882
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,8,32,power_law_1.2,0.1228876781463623
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,8,32,power_law_1.2,0.12624256134033202
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,8,32,power_law_1.2,0.12880640029907225
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,8,32,power_law_1.2,0.13687680244445802
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,8,32,power_law_1.2,0.06972288131713868
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,8,32,power_law_1.2,0.13260416030883787
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,8,32,power_law_1.2,0.14012160301208496
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,8,32,power_law_1.2,0.15545344352722168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,8,32,power_law_1.2,0.1656755256652832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,8,32,power_law_1.2,0.1982464027404785
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,8,32,power_law_1.2,0.1176307201385498
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,8,32,power_law_1.01,0.06930560111999512
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,8,32,power_law_1.2,0.2950758361816406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,8,32,power_law_1.2,0.3734873580932617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,8,32,power_law_1.2,0.07128575801849366
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,8,32,power_law_1.2,0.48303871154785155
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,8,32,power_law_1.2,0.13309823989868164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,8,32,power_law_1.2,0.040252161026000974
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,8,32,power_law_1.01,0.8449100494384766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,8,32,power_law_1.2,0.627979507446289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,8,32,power_law_1.2,0.04837759971618653
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,8,32,power_law_1.2,0.048428797721862794
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,8,32,power_law_1.2,0.22114816665649414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,8,32,power_law_1.2,0.050343680381774905
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,8,32,power_law_1.2,0.06569983959197997
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,8,32,power_law_1.2,0.9887334442138671
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,8,32,power_law_1.2,0.06762495994567871
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,8,32,power_law_1.2,0.06900991916656493
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,8,32,power_law_1.2,0.07196928024291992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,8,32,power_law_1.2,0.07101568222045898
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,8,32,power_law_1.2,0.07334400177001953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,8,32,power_law_1.2,0.07608960151672364
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,8,32,power_law_1.2,0.08261119842529296
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,8,32,power_law_1.2,0.07868800163269044
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,8,32,power_law_1.2,1.4691212463378907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,8,32,power_law_1.2,0.07942399978637696
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,8,32,power_law_1.2,0.08412544250488281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,8,32,power_law_1.2,0.09350912094116211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,8,32,power_law_1.2,0.11198719978332519
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,8,32,power_law_1.2,0.05100287914276123
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,8,32,power_law_1.2,0.14376832008361817
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,8,32,power_law_1.2,0.1684671974182129
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,8,32,power_law_1.2,0.21353343963623045
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,8,32,power_law_1.2,0.27399679183959963
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,16,16,balanced,0.03046655893325806
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,16,16,balanced,0.0303436803817749
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,8,32,power_law_1.2,0.39766654968261717
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,8,32,power_law_1.2,0.4668505477905273
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,8,32,power_law_1.2,0.08019840240478515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,16,16,balanced,0.03489408016204834
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,16,16,balanced,0.05835391998291015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,16,16,balanced,0.06745215892791748
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,16,16,balanced,0.12272640228271485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,16,16,balanced,0.1252953624725342
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,16,16,balanced,0.12968064308166505
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,8,32,power_law_1.2,0.8759027099609374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,16,16,balanced,0.135729923248291
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,16,16,balanced,0.13077376365661622
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,8,32,power_law_1.2,1.1071782684326172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,16,16,balanced,0.13606528282165525
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,16,16,balanced,0.13661312103271483
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,16,16,balanced,0.14246272087097167
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,16,16,balanced,0.1624985694885254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,16,16,balanced,0.18149120330810548
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,16,16,balanced,0.031357440948486325
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,16,16,balanced,0.20240768432617187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,16,16,balanced,0.24285184860229495
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,16,16,balanced,0.07757184028625488
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,16,16,balanced,0.36957569122314454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,16,16,balanced,0.12487680435180663
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,16,16,balanced,0.13363327980041503
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,16,16,balanced,0.45848831176757815
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,16,16,balanced,0.0304256010055542
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,16,16,balanced,0.0303603196144104
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,16,16,balanced,0.03082880020141602
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,16,16,balanced,0.03333631992340088
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,16,16,balanced,0.03672192096710205
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,16,16,balanced,0.6317824172973633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,16,16,balanced,0.05026559829711914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,16,16,balanced,0.06038271903991699
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,16,16,balanced,0.08435199737548829
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,16,16,balanced,0.15261311531066896
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,16,16,balanced,0.08607232093811035
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,16,16,balanced,0.0919206428527832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,16,16,balanced,0.09159680366516113
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,16,16,balanced,0.8465535736083984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,16,16,balanced,0.08809984207153321
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,16,16,balanced,0.09048959732055664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,16,16,balanced,0.09155455589294434
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,16,16,balanced,0.09515904426574708
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,16,16,balanced,0.10489727973937987
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,16,16,balanced,0.11665663719177247
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,16,16,balanced,0.1367628765106201
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,16,16,balanced,0.1569919967651367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,16,16,balanced,0.19713279724121094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,16,16,balanced,0.23793535232543944
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,16,16,balanced,0.3141529655456543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,16,16,power_law_1.01,0.05546751976013183
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,16,16,balanced,0.4163942337036133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,16,16,power_law_1.01,0.06455423831939697
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,16,16,balanced,0.08586112022399903
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,16,16,balanced,0.5717273712158203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,16,16,balanced,0.08992639541625977
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,16,16,power_law_1.01,0.07154816150665283
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,16,16,power_law_1.01,0.07223680019378662
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,16,16,balanced,0.7511820983886719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,16,16,power_law_1.01,0.11846400260925294
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,16,16,balanced,0.2898496055603027
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,16,16,power_law_1.01,0.11816960334777833
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,16,16,power_law_1.01,0.11844608306884766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,16,16,power_law_1.01,0.11963264465332031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,16,16,power_law_1.01,0.12101119995117186
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,16,16,power_law_1.01,0.1296243190765381
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,16,16,power_law_1.01,0.12927488327026365
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,16,16,power_law_1.01,0.13122943878173826
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,16,16,power_law_1.01,0.13359487533569336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,16,16,power_law_1.01,0.13754624366760254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,16,16,power_law_1.01,0.14240896224975585
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,16,16,power_law_1.01,0.15652735710144045
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,16,16,power_law_1.01,0.17054079055786134
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,16,16,power_law_1.01,0.19883647918701172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,16,16,power_law_1.01,0.22813184738159178
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,16,16,power_law_1.01,0.303824634552002
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,16,16,power_law_1.01,0.07733503818511964
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,16,16,power_law_1.01,0.3653363037109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,16,16,power_law_1.01,0.4731238555908203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,16,16,power_law_1.01,0.034967041015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,16,16,power_law_1.01,0.5531622314453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,16,16,power_law_1.01,0.04692224025726318
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,16,16,power_law_1.01,0.1289344024658203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,16,16,power_law_1.01,0.05760000228881836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,16,16,power_law_1.01,0.05606143951416016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,16,16,power_law_1.01,0.908089599609375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,16,16,power_law_1.01,0.0821235179901123
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,16,16,power_law_1.01,0.08217215538024902
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,16,16,power_law_1.01,0.08273792266845703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,16,16,power_law_1.01,0.0818995189666748
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,16,16,power_law_1.01,0.08742783546447755
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,16,16,power_law_1.01,0.09192831993103028
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,16,16,power_law_1.01,1.1201036834716798
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,16,16,power_law_1.01,0.09450495719909668
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,16,16,power_law_1.01,0.08504704475402833
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,16,16,power_law_1.01,0.08944512367248535
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,16,16,power_law_1.01,0.09172479629516601
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,16,16,power_law_1.01,0.09610367774963378
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,16,16,power_law_1.01,0.11060864448547363
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,16,16,power_law_1.01,0.12255999565124512
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,16,16,power_law_1.01,0.1488371181488037
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,16,16,power_law_1.01,0.05830656051635742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,16,16,power_law_1.01,0.23820159912109373
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,16,16,power_law_1.01,0.27601919174194334
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,16,16,power_law_1.01,0.07957759857177735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,16,16,power_law_1.2,0.05504000186920166
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,16,16,power_law_1.01,0.5394227218627929
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,16,16,power_law_1.2,0.0644595193862915
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,16,16,power_law_1.2,0.06899456024169923
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,16,16,power_law_1.2,0.07360000133514404
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,16,16,power_law_1.01,0.7486579132080078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,16,16,power_law_1.2,0.07462399959564209
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,16,16,power_law_1.01,0.1748236846923828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,16,16,power_law_1.2,0.1206105613708496
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,16,16,power_law_1.2,0.11928959846496583
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,16,16,power_law_1.01,0.9369331359863281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,16,16,power_law_1.2,0.12087295532226563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,16,16,power_law_1.2,0.12146431922912597
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,16,16,power_law_1.2,0.12756735801696778
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,16,16,power_law_1.2,0.1303910446166992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,16,16,power_law_1.2,0.13496319770812987
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,16,16,power_law_1.2,0.1307033634185791
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,16,16,power_law_1.01,0.4232230377197266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,16,16,power_law_1.2,0.13450112342834472
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,16,16,power_law_1.2,0.1385548782348633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,16,16,power_law_1.2,0.14153599739074707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,16,16,power_law_1.2,0.15888511657714843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,16,16,power_law_1.2,0.17811712265014648
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,16,16,power_law_1.2,0.21858432769775388
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,16,16,power_law_1.2,0.2452249526977539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,16,16,power_law_1.2,0.3109875106811523
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,16,16,power_law_1.2,0.39492095947265626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,16,16,power_law_1.2,0.03533695936203003
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,16,16,power_law_1.2,0.11869312286376954
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,16,16,power_law_1.2,0.5235558319091796
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,16,16,power_law_1.2,0.04661759853363037
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,16,16,power_law_1.2,0.053267202377319335
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,16,16,power_law_1.2,0.055973119735717766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,16,16,power_law_1.2,0.05901951789855957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,16,16,power_law_1.2,0.8191206359863281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,16,16,power_law_1.2,0.08222463607788086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,16,16,power_law_1.2,0.07536511898040772
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,16,16,power_law_1.2,0.08271103858947754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,16,16,power_law_1.2,0.0822003173828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,16,16,power_law_1.2,1.0913996887207031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,16,16,power_law_1.2,0.09049216270446778
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,16,16,power_law_1.2,0.08458239555358886
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,16,16,power_law_1.2,0.08762751579284668
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,16,16,power_law_1.2,0.08816896438598633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,16,16,power_law_1.2,0.09302783966064453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,16,16,power_law_1.2,0.09819775581359863
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,16,16,power_law_1.2,0.11792896270751954
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,16,16,power_law_1.2,1.568554229736328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,16,16,power_law_1.2,0.13603967666625977
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,16,16,power_law_1.2,0.16487039566040038
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,16,16,power_law_1.2,0.2014835166931152
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,16,16,power_law_1.2,0.2510745620727539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,16,16,power_law_1.2,0.3295539093017578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,32,8,balanced,0.031239678859710695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,32,8,balanced,0.03302016019821167
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,16,16,power_law_1.2,0.44707073211669923
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,32,8,balanced,0.03670016050338745
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,16,16,power_law_1.2,0.09237248420715331
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,16,16,power_law_1.2,0.6057855987548828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,32,8,balanced,0.04299263954162598
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,32,8,balanced,0.07928832054138184
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,32,8,balanced,0.09080831527709961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,32,8,balanced,0.13203840255737304
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,32,8,balanced,0.13501567840576173
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,32,8,balanced,0.13761407852172852
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,32,8,balanced,0.13182463645935058
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,32,8,balanced,0.1355776023864746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,32,8,balanced,0.13837823867797852
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,32,8,balanced,0.1455014419555664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,16,16,power_law_1.2,1.1313855743408203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,32,8,balanced,0.15041152000427246
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,32,8,balanced,0.15509887695312502
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,32,8,balanced,0.1623040008544922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,32,8,balanced,0.1818956756591797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,32,8,balanced,0.20046335220336914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,16,16,power_law_1.2,0.0806937599182129
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,32,8,balanced,0.24040191650390624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,32,8,balanced,0.28041088104248046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,32,8,balanced,0.3615244674682617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,32,8,balanced,0.4505164718627929
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,16,16,power_law_1.2,0.9345331573486328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,32,8,balanced,0.6094323348999023
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,32,8,balanced,0.03201152086257934
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,32,8,balanced,0.032529919147491454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,32,8,balanced,0.040110077857971195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,32,8,balanced,0.048463358879089355
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,32,8,balanced,0.06604544162750245
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,32,8,balanced,0.08353023529052735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,32,8,balanced,1.1087283325195312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,32,8,balanced,0.06628352165222168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,32,8,balanced,0.11172608375549316
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,32,8,balanced,0.10950655937194824
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,32,8,balanced,0.11421567916870116
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,32,8,balanced,0.11329407691955566
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,32,8,balanced,0.11475711822509767
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,32,8,balanced,0.12056575775146486
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,32,8,balanced,1.4904588317871093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,32,8,balanced,0.1298636817932129
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,32,8,balanced,0.1391487979888916
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,32,8,balanced,0.7779507446289062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,32,8,balanced,0.1786675262451172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,32,8,balanced,0.2184524726867676
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,32,8,balanced,0.2552716827392578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,32,8,balanced,0.33303680419921877
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,32,8,balanced,0.41088897705078126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,32,8,balanced,0.5658841705322265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,32,8,balanced,0.1238912010192871
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,32,8,balanced,0.7617727661132813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,32,8,power_law_1.01,0.035073280334472656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,32,8,balanced,0.035569920539855956
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,32,8,power_law_1.01,0.05842432022094727
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,32,8,power_law_1.01,0.0704588794708252
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,32,8,power_law_1.01,0.07086080074310303
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,32,8,power_law_1.01,0.07647488117218018
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,32,8,balanced,1.0740172576904297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,32,8,power_law_1.01,0.08388480186462402
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,32,8,balanced,0.11085184097290039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,32,8,power_law_1.01,0.08841728210449219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,32,8,power_law_1.01,0.1302143955230713
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,32,8,power_law_1.01,0.12980992317199708
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,32,8,power_law_1.01,0.13201536178588866
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,32,8,power_law_1.01,0.1320678424835205
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,32,8,balanced,1.4645555114746094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,32,8,power_law_1.01,0.13465215682983397
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,32,8,power_law_1.01,0.14001279830932617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,32,8,power_law_1.01,0.14308735847473145
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,32,8,power_law_1.01,0.15256832122802735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,32,8,balanced,0.1576959991455078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,32,8,power_law_1.01,0.1645132827758789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,32,8,power_law_1.01,0.1851353645324707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,32,8,power_law_1.01,0.2176243209838867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,32,8,power_law_1.01,0.2543296051025391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,32,8,power_law_1.01,0.32871551513671876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,32,8,power_law_1.01,0.39488384246826175
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,32,8,power_law_1.01,0.0366425609588623
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,32,8,power_law_1.01,0.13211520195007326
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,32,8,power_law_1.01,0.5401587295532226
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,32,8,power_law_1.01,0.04274816036224365
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,32,8,power_law_1.01,0.7065907287597656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,32,8,power_law_1.01,0.058142719268798826
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,32,8,power_law_1.01,0.05659008026123047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,32,8,power_law_1.01,0.06885503768920899
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,32,8,power_law_1.01,0.07658368110656738
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,32,8,power_law_1.01,0.9805145263671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,32,8,power_law_1.01,0.07976704120635987
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,32,8,power_law_1.01,0.1055513572692871
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,32,8,power_law_1.01,0.10753919601440429
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,32,8,power_law_1.01,0.11131648063659667
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,32,8,power_law_1.01,0.1069606399536133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,32,8,power_law_1.01,0.10651264190673829
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,32,8,power_law_1.01,0.10736895561218263
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,32,8,power_law_1.01,1.3842880249023437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,32,8,power_law_1.01,0.11314047813415526
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,32,8,power_law_1.01,0.11709440231323243
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,32,8,power_law_1.01,0.11853952407836914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,32,8,power_law_1.01,0.1314086437225342
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,32,8,power_law_1.01,0.13843071937561036
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,32,8,power_law_1.01,0.19397375106811524
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,32,8,power_law_1.01,0.22917375564575196
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,32,8,power_law_1.01,0.29843711853027344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,32,8,power_law_1.01,0.37750270843505856
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,32,8,power_law_1.01,0.10292096138000488
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,32,8,power_law_1.01,0.5241958236694335
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,32,8,power_law_1.2,0.034816000461578366
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,32,8,power_law_1.2,0.05894271850585937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,32,8,power_law_1.2,0.06505343914031983
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,32,8,power_law_1.2,0.07054080009460449
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,32,8,power_law_1.01,0.1599961566925049
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,32,8,power_law_1.2,0.07604991912841796
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,32,8,power_law_1.01,1.0229977416992189
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,32,8,power_law_1.2,0.08355711936950684
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,32,8,power_law_1.2,0.08837247848510742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,32,8,power_law_1.2,0.12747520446777344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,32,8,power_law_1.01,1.373175048828125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,32,8,power_law_1.2,0.1314291191101074
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,32,8,power_law_1.2,0.13253376007080078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,32,8,power_law_1.2,0.1303667163848877
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,32,8,power_law_1.01,0.6828313446044921
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,32,8,power_law_1.2,0.13935999870300292
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,32,8,power_law_1.2,0.14547200202941896
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,32,8,power_law_1.2,0.15149824142456053
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,32,8,power_law_1.2,0.1697702407836914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,32,8,power_law_1.2,0.1886720085144043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,32,8,power_law_1.2,0.22451967239379883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,32,8,power_law_1.2,0.2604198455810547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,32,8,power_law_1.2,0.12794367790222166
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,32,8,power_law_1.2,0.33732864379882815
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,32,8,power_law_1.2,0.13179648399353028
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,32,8,power_law_1.2,0.4169023895263672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,32,8,power_law_1.2,0.03671040058135987
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,32,8,power_law_1.2,0.5470899200439453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,32,8,power_law_1.2,0.13566335678100586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,32,8,power_law_1.2,0.733148193359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,32,8,power_law_1.2,0.051842560768127446
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,32,8,power_law_1.2,0.0566374397277832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,32,8,power_law_1.2,0.06517632007598877
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,32,8,power_law_1.2,0.07930111885070801
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,32,8,power_law_1.2,0.10488320350646974
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,32,8,power_law_1.2,0.10344320297241212
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,32,8,power_law_1.2,0.10091648101806641
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,32,8,power_law_1.2,0.10835968017578126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,32,8,power_law_1.2,0.10422911643981933
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,32,8,power_law_1.2,0.10887040138244627
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,32,8,power_law_1.2,1.4361740112304688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,32,8,power_law_1.2,0.11265407562255861
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,32,8,power_law_1.2,0.042615041732788086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,32,8,power_law_1.2,0.11728639602661133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,32,8,power_law_1.2,0.1225267219543457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,32,8,power_law_1.2,0.16221952438354492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,32,8,power_law_1.2,0.15074432373046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,32,8,power_law_1.2,0.20728191375732422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,32,8,power_law_1.2,0.24108160018920896
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,32,8,power_law_1.2,0.07935232162475586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,32,8,power_law_1.2,1.1307161712646485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,32,8,power_law_1.2,0.3258393478393555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,1,256,balanced,0.022914559841156007
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,32,8,power_law_1.2,0.38996734619140627
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,1,256,balanced,0.022909440994262696
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,32,8,power_law_1.2,0.10794367790222167
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,1,256,balanced,0.022938880920410156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,1,256,balanced,0.022853119373321535
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,1,256,balanced,0.023046400547027588
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,32,8,power_law_1.2,0.5699571228027344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,1,256,balanced,0.022740480899810792
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,1,256,balanced,0.03920000076293945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,1,256,balanced,0.039571199417114254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,1,256,balanced,0.03969536066055298
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,1,256,balanced,0.039585280418396
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,1,256,balanced,0.0397273588180542
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,1,256,balanced,0.03997695922851562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,1,256,balanced,0.04010496139526367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,1,256,balanced,0.03998591899871826
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,1,256,balanced,0.04026368141174316
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,1,256,balanced,0.041103358268737795
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,1,256,balanced,0.040643839836120604
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,1,256,balanced,0.05740799903869629
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,1,256,balanced,0.041621761322021486
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,1,256,balanced,0.05068927764892578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,1,256,balanced,0.05074687957763672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,32,8,power_law_1.2,1.048716812133789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,1,256,balanced,0.06542848110198975
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,1,256,balanced,0.07286272048950196
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,1,256,balanced,0.08053248405456544
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,1,256,balanced,0.09937664031982421
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,1,256,balanced,0.024853758811950684
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,1,256,balanced,0.024734721183776856
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,1,256,balanced,0.024883201122283934
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,32,8,power_law_1.2,1.356299591064453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,1,256,balanced,0.024849920272827147
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,1,256,balanced,0.024401919841766356
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,1,256,balanced,0.025085439682006837
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,1,256,balanced,0.03909375905990601
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,1,256,balanced,0.039146239757537844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,1,256,balanced,0.03868544101715088
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,1,256,balanced,0.03939199924468994
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,1,256,balanced,0.03906816005706787
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,1,256,balanced,0.039548161029815676
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,32,8,power_law_1.2,0.7206374359130859
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,1,256,balanced,0.03992063999176025
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,1,256,balanced,0.039598081111907955
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,1,256,balanced,0.04069759845733643
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,1,256,balanced,0.0399014401435852
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,1,256,balanced,0.040643839836120604
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,1,256,balanced,0.043333120346069336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,1,256,balanced,0.04495872020721435
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,1,256,balanced,0.04889599800109863
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,1,256,balanced,0.06723711967468263
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,1,256,balanced,0.046291198730468754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,1,256,balanced,0.053116159439086916
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,1,256,balanced,0.05743103981018066
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,1,256,balanced,0.0648524808883667
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,1,256,balanced,0.07362304210662843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,1,256,balanced,0.10151167869567872
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,1,256,power_law_1.01,0.022682878971099853
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,1,256,power_law_1.01,0.022714879512786865
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,1,256,power_law_1.01,0.02345344066619873
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,1,256,power_law_1.01,0.02268928050994873
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,1,256,power_law_1.01,0.039448320865631104
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,1,256,power_law_1.01,0.03987711906433106
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,1,256,power_law_1.01,0.04009088039398193
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,1,256,power_law_1.01,0.04362751960754395
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,1,256,power_law_1.01,0.04068480014801025
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,1,256,power_law_1.01,0.040270080566406255
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,1,256,power_law_1.01,0.048565759658813476
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,1,256,balanced,0.04127103805541992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,1,256,power_law_1.01,0.05246592044830323
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,1,256,power_law_1.01,0.05049088001251221
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,1,256,power_law_1.01,0.0429747200012207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,1,256,power_law_1.01,0.06362239837646484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,1,256,power_law_1.01,0.06884479999542237
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,1,256,power_law_1.01,0.06920447826385498
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,1,256,power_law_1.01,0.0781440019607544
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,1,256,power_law_1.01,0.09166848182678222
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,1,256,power_law_1.01,0.10231040000915528
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,1,256,power_law_1.01,0.1574451160430908
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,1,256,power_law_1.01,0.1894822311401367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,1,256,power_law_1.01,0.2728217506408691
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,1,256,power_law_1.01,0.35871871948242184
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,1,256,power_law_1.01,0.024892160892486574
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,1,256,power_law_1.01,0.04083199977874756
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,1,256,power_law_1.01,0.024611840248107912
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,1,256,power_law_1.01,0.0245580792427063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,1,256,power_law_1.01,0.552863998413086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,1,256,balanced,0.04309887886047363
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,1,256,power_law_1.01,0.03896575927734375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,1,256,power_law_1.01,0.039690239429473875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,1,256,power_law_1.01,0.039377920627593994
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,1,256,power_law_1.01,0.039823360443115234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,1,256,power_law_1.01,0.040849919319152835
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,1,256,power_law_1.01,0.04030208110809326
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,1,256,power_law_1.01,0.04131328105926514
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,1,256,power_law_1.01,0.042548480033874514
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,1,256,power_law_1.01,0.04209792137145996
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,1,256,power_law_1.01,0.043701758384704595
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,1,256,power_law_1.01,0.044906239509582516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,1,256,power_law_1.01,0.04888576030731201
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,1,256,power_law_1.01,0.04943871974945068
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,1,256,power_law_1.01,0.05462656021118164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,1,256,power_law_1.01,0.06265727996826172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,1,256,power_law_1.01,0.07039872169494629
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,1,256,power_law_1.01,0.08900863647460937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,1,256,power_law_1.01,0.11341312408447266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,1,256,power_law_1.01,0.16339967727661134
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,1,256,power_law_1.01,0.20757888793945312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,1,256,power_law_1.01,0.7029004669189453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,1,256,power_law_1.01,0.314391040802002
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,1,256,power_law_1.2,0.022725119590759277
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,1,256,power_law_1.2,0.023371520042419432
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,1,256,power_law_1.01,0.44860671997070317
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,1,256,power_law_1.2,0.03908992052078247
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,1,256,power_law_1.2,0.03938047885894776
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,1,256,power_law_1.01,0.041463041305541994
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,1,256,power_law_1.2,0.03984127998352051
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,1,256,power_law_1.2,0.03984512090682983
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,1,256,power_law_1.2,0.04065408229827881
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,1,256,power_law_1.2,0.046936321258544925
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,1,256,power_law_1.2,0.04689792156219482
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,1,256,power_law_1.2,0.054374399185180666
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,1,256,power_law_1.2,0.05305471897125245
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,1,256,power_law_1.2,0.06660863876342774
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,1,256,power_law_1.2,0.07050496101379394
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,1,256,power_law_1.01,0.02471679925918579
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,1,256,power_law_1.2,0.07311744213104247
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,1,256,power_law_1.2,0.08031359672546387
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,1,256,power_law_1.2,0.09175807952880859
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,1,256,power_law_1.2,0.11397888183593749
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,1,256,power_law_1.2,0.1536678409576416
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,1,256,power_law_1.2,0.21822208404541016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,1,256,power_law_1.2,0.33179134368896485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,1,256,power_law_1.2,0.43216384887695314
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,1,256,power_law_1.2,0.043148798942565916
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,1,256,power_law_1.2,0.02483328104019165
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,1,256,power_law_1.2,0.05392000198364257
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,1,256,power_law_1.2,0.661612777709961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,1,256,power_law_1.2,0.02487936019897461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,1,256,power_law_1.2,0.02473344087600708
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,1,256,power_law_1.2,0.03923072099685669
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,1,256,power_law_1.2,0.03901695966720581
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,1,256,power_law_1.2,0.03947904109954834
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,1,256,power_law_1.2,0.0395251202583313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,1,256,power_law_1.2,0.023120639324188234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,1,256,power_law_1.2,0.04034560203552246
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,1,256,power_law_1.2,0.04024831771850586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,1,256,power_law_1.2,0.04061183929443359
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,1,256,power_law_1.2,1.1331903839111328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,1,256,power_law_1.2,0.04124927997589111
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,1,256,power_law_1.2,0.043955202102661135
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,1,256,power_law_1.2,0.04595839977264404
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,1,256,power_law_1.2,0.04943999767303467
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,1,256,power_law_1.2,0.04974463939666748
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,1,256,power_law_1.2,0.05564032077789307
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,1,256,power_law_1.2,0.06263679981231689
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,1,256,power_law_1.2,1.5844863891601562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,1,256,power_law_1.2,0.07556479930877685
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,1,256,power_law_1.2,0.09317888259887695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,1,256,power_law_1.2,0.12927359580993653
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,1,256,power_law_1.2,0.18779008865356445
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,1,256,power_law_1.2,0.25943296432495117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,2,128,balanced,0.020725760459899902
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,1,256,power_law_1.2,0.3945753479003906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,2,128,balanced,0.020360960960388183
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,2,128,balanced,0.020543999671936035
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,2,128,balanced,0.020546560287475583
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,1,256,power_law_1.2,0.04256768226623535
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,2,128,balanced,0.02145535945892334
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,2,128,balanced,0.03055232048034668
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,2,128,balanced,0.03078655958175659
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,2,128,balanced,0.03105407953262329
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,2,128,balanced,0.031128320693969723
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,2,128,balanced,0.031568639278411866
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,2,128,balanced,0.0315391993522644
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,1,256,power_law_1.2,0.9079360198974609
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,2,128,balanced,0.03171583890914917
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,2,128,balanced,0.033054718971252436
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,2,128,balanced,0.03817343950271607
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,2,128,balanced,0.03728384017944336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,2,128,balanced,0.03877631902694702
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,2,128,balanced,0.04229248046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,2,128,balanced,0.04772480010986328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,2,128,balanced,0.06043007850646973
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,2,128,balanced,0.0564467191696167
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,2,128,balanced,0.06447743892669677
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,2,128,balanced,0.06895743846893311
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,2,128,balanced,0.07654528141021728
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,2,128,balanced,0.08330623626708984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,2,128,balanced,0.10579327583312988
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,1,256,power_law_1.2,0.042954239845275875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,2,128,balanced,0.12090623855590821
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,2,128,balanced,0.02266239881515503
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,1,256,power_law_1.2,0.6483929443359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,2,128,balanced,0.022625279426574708
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,2,128,balanced,0.022763519287109374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,2,128,balanced,0.02283008098602295
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,2,128,balanced,0.02323199987411499
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,2,128,balanced,0.03546495914459229
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,2,128,balanced,0.03548160076141357
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,2,128,balanced,0.036110079288482665
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,2,128,balanced,0.03590271949768066
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,2,128,balanced,0.036215040683746334
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,2,128,balanced,0.03636224031448364
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,2,128,balanced,0.03751552104949951
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,2,128,balanced,0.03880320072174072
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,2,128,balanced,0.03753599882125854
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,2,128,balanced,0.038279678821563724
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,2,128,balanced,0.039057919979095454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,2,128,balanced,0.04090496063232422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,2,128,balanced,0.04332159996032715
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,2,128,balanced,0.04023551940917969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,2,128,balanced,0.04662655830383301
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,2,128,balanced,0.049976320266723634
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,2,128,balanced,0.056985602378845215
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,2,128,balanced,0.06254591941833496
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,2,128,balanced,0.07742720127105714
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,2,128,balanced,0.09368191719055176
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,2,128,power_law_1.01,0.02131839990615845
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,2,128,balanced,0.020477440357208252
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,2,128,power_law_1.01,0.02152064085006714
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,2,128,power_law_1.01,0.02158207893371582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,2,128,power_law_1.01,0.02151423931121826
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,2,128,power_law_1.01,0.03075968027114868
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,2,128,power_law_1.01,0.03184767961502075
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,2,128,power_law_1.01,0.03278847932815552
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,2,128,power_law_1.01,0.037205760478973386
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,2,128,power_law_1.01,0.03893887996673584
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,2,128,balanced,0.03580543994903564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,2,128,power_law_1.01,0.04428544044494629
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,2,128,power_law_1.01,0.04280831813812256
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,2,128,power_law_1.01,0.04773248195648193
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,2,128,power_law_1.01,0.0542195177078247
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,2,128,power_law_1.01,0.059399681091308595
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,2,128,power_law_1.01,0.061743359565734866
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,2,128,power_law_1.01,0.06567296028137207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,2,128,power_law_1.01,0.07273087978363038
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,2,128,power_law_1.01,0.0839577579498291
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,2,128,power_law_1.01,0.10006272315979003
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,2,128,power_law_1.01,0.1363481616973877
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,2,128,power_law_1.01,0.14761728286743164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,2,128,power_law_1.01,0.2395302391052246
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,2,128,balanced,0.022887680530548096
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,2,128,power_law_1.01,0.02280832052230835
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,2,128,power_law_1.01,0.031374080181121825
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,2,128,power_law_1.01,0.023184640407562258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,2,128,power_law_1.01,0.4589862442016601
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,2,128,power_law_1.01,0.02288383960723877
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,2,128,power_law_1.01,0.5877721786499024
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,2,128,power_law_1.01,0.023054080009460447
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,2,128,power_law_1.01,0.035665919780731206
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,2,128,power_law_1.01,0.03605376005172729
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,2,128,power_law_1.01,0.06286975860595703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,2,128,power_law_1.01,0.035980799198150636
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,2,128,power_law_1.01,0.03671551942825317
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,2,128,power_law_1.01,0.038054399490356446
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,2,128,power_law_1.01,0.040586237907409665
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,2,128,power_law_1.01,0.04236927986145019
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,2,128,power_law_1.01,0.2907571220397949
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,2,128,power_law_1.01,0.04385663986206055
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,2,128,power_law_1.01,0.04603263854980469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,2,128,power_law_1.01,0.047411198616027835
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,2,128,power_law_1.01,0.05268095970153809
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,2,128,power_law_1.01,0.06083072185516357
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,2,128,power_law_1.01,0.06532608032226563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,2,128,power_law_1.01,0.09918463706970215
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,2,128,power_law_1.01,0.036528639793395996
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,2,128,power_law_1.01,0.14969599723815918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,2,128,power_law_1.01,0.03852031946182251
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,2,128,power_law_1.01,0.19711360931396485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,2,128,power_law_1.2,0.021615359783172607
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,2,128,power_law_1.01,0.2677926445007324
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,2,128,power_law_1.2,0.021568000316619873
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,2,128,power_law_1.01,0.35964160919189453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,2,128,power_law_1.2,0.021477119922637938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,2,128,power_law_1.2,0.030551040172576906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,2,128,power_law_1.2,0.030936319828033448
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,2,128,power_law_1.2,0.0323801589012146
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,2,128,power_law_1.2,0.031741440296173096
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,2,128,power_law_1.2,0.03330559968948364
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,2,128,power_law_1.2,0.04152959823608399
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,2,128,power_law_1.01,0.11075072288513184
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,2,128,power_law_1.2,0.05237247943878174
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,2,128,power_law_1.2,0.03918976068496704
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,2,128,power_law_1.2,0.05342847824096679
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,2,128,power_law_1.01,0.03888256072998047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,2,128,power_law_1.2,0.06303359985351562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,2,128,power_law_1.2,0.06214911937713623
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,2,128,power_law_1.2,0.06260223865509033
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,2,128,power_law_1.2,0.07390079975128175
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,2,128,power_law_1.2,0.07965184211730957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,2,128,power_law_1.2,0.09585023880004882
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,2,128,power_law_1.2,0.12243968009948732
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,2,128,power_law_1.2,0.16596479415893556
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,2,128,power_law_1.2,0.2332159996032715
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,2,128,power_law_1.2,0.34011905670166015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,2,128,power_law_1.2,0.02290560007095337
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,2,128,power_law_1.2,0.5093081665039063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,2,128,power_law_1.2,0.04403071880340576
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,2,128,power_law_1.2,0.023152639865875242
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,2,128,power_law_1.2,0.02286207914352417
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,2,128,power_law_1.01,0.037978880405426026
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,2,128,power_law_1.2,0.03580032110214233
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,2,128,power_law_1.2,0.786434555053711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,2,128,power_law_1.2,0.03574527978897095
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,2,128,power_law_1.2,0.036145920753479006
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,2,128,power_law_1.2,0.03624576091766357
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,2,128,power_law_1.2,0.03665152072906494
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,2,128,power_law_1.2,0.03839359998703003
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,2,128,power_law_1.2,0.03821439981460571
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,2,128,power_law_1.2,0.038526721000671386
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,2,128,power_law_1.2,0.03972608089447022
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,2,128,power_law_1.2,0.04209536075592041
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,2,128,power_law_1.2,0.04329599857330323
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,2,128,power_law_1.2,0.04488704204559326
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,2,128,power_law_1.2,0.04743936061859131
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,2,128,power_law_1.2,0.05115903854370117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,2,128,power_law_1.2,0.057299199104309084
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,2,128,power_law_1.2,0.0661299180984497
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,2,128,power_law_1.2,0.08383359909057617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,2,128,power_law_1.2,0.11849599838256837
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,2,128,power_law_1.2,0.15480704307556153
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,2,128,power_law_1.2,0.05917952060699463
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,2,128,power_law_1.2,0.23575551986694335
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,4,64,balanced,0.019485440254211426
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,4,64,balanced,0.019281920194625855
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,4,64,balanced,0.019073280096054076
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,2,128,power_law_1.2,0.35979007720947265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,4,64,balanced,0.0189683198928833
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,4,64,balanced,0.019952640533447266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,4,64,balanced,0.02111999988555908
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,4,64,balanced,0.031116800308227537
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,4,64,balanced,0.035929598808288575
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,2,128,power_law_1.2,0.038982400894165034
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,2,128,power_law_1.2,0.5293862533569336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,4,64,balanced,0.030420479774475095
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,4,64,balanced,0.031768319606781004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,4,64,balanced,0.034403839111328126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,4,64,balanced,0.03131776094436646
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,4,64,balanced,0.03284735918045044
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,2,128,power_law_1.2,0.7204134368896484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,4,64,balanced,0.039874560832977295
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,4,64,balanced,0.04048255920410156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,4,64,balanced,0.05516416072845459
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,4,64,balanced,0.05306623935699463
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,4,64,balanced,0.06400000095367432
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,4,64,balanced,0.07065087795257569
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,4,64,balanced,0.07655039787292481
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,4,64,balanced,0.09112575531005859
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,4,64,balanced,0.10347904205322264
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,4,64,balanced,0.13507583618164062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,4,64,balanced,0.021612799167633055
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,4,64,balanced,0.16221952438354492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,4,64,balanced,0.02184319972991943
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,4,64,balanced,0.021684479713439942
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,2,128,power_law_1.2,1.1371532440185548
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,4,64,balanced,0.02220671892166138
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,4,64,balanced,0.022530560493469236
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,4,64,balanced,0.033960959911346435
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,4,64,balanced,0.03410943984985352
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,4,64,balanced,0.034485759735107424
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,4,64,balanced,0.03441152095794678
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,4,64,balanced,0.03500288009643555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,4,64,balanced,0.05793536186218262
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,4,64,balanced,0.03546623945236206
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,4,64,balanced,0.03557503938674927
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,4,64,balanced,0.03663615942001343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,4,64,balanced,0.03712127923965454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,4,64,balanced,0.03841279983520508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,4,64,balanced,0.04107647895812988
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,4,64,balanced,0.03817728042602539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,4,64,balanced,0.04278272151947021
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,4,64,balanced,0.04616191864013672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,4,64,balanced,0.05157887935638428
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,4,64,balanced,0.03017215967178345
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,4,64,balanced,0.05763967990875244
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,4,64,balanced,0.07233280181884766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,4,64,balanced,0.08567040443420411
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,4,64,balanced,0.11274880409240721
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,4,64,balanced,0.03184639930725098
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,4,64,balanced,0.13721343994140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,4,64,power_law_1.01,0.02061311960220337
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,4,64,power_law_1.01,0.021386239528656006
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,4,64,power_law_1.01,0.02113663911819458
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,4,64,power_law_1.01,0.020826880931854246
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,4,64,power_law_1.01,0.03336191892623901
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,4,64,power_law_1.01,0.030191359519958494
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,4,64,balanced,0.03655807971954346
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,4,64,power_law_1.01,0.030635519027709963
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,4,64,power_law_1.01,0.04644735813140869
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,4,64,power_law_1.01,0.045154561996459966
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,4,64,power_law_1.01,0.03607295989990235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,4,64,power_law_1.01,0.034961919784545895
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,4,64,power_law_1.01,0.04863232135772705
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,4,64,power_law_1.01,0.05477248191833496
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,4,64,power_law_1.01,0.056880640983581546
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,4,64,power_law_1.01,0.05993855953216552
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,4,64,power_law_1.01,0.0561625623703003
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,4,64,balanced,0.021553919315338135
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,4,64,power_law_1.01,0.06320511817932128
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,4,64,power_law_1.01,0.06914432048797607
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,4,64,power_law_1.01,0.08006912231445312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,4,64,power_law_1.01,0.09477760314941405
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,4,64,power_law_1.01,0.1211903953552246
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,4,64,power_law_1.01,0.15104255676269532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,4,64,power_law_1.01,0.19603967666625977
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,4,64,power_law_1.01,0.022498559951782227
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,4,64,power_law_1.01,0.2997952079772949
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,4,64,power_law_1.01,0.022269439697265626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,4,64,power_law_1.01,0.022455039024353026
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,4,64,power_law_1.01,0.4584128189086914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,4,64,power_law_1.01,0.022382080554962158
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,4,64,power_law_1.01,0.03393663883209229
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,4,64,power_law_1.01,0.5344153594970703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,4,64,power_law_1.01,0.034599680900573734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,4,64,power_law_1.01,0.035352320671081544
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,4,64,power_law_1.01,0.03594624042510987
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,4,64,power_law_1.01,0.03615103960037232
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,4,64,power_law_1.01,0.03668224096298218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,4,64,power_law_1.01,0.037611520290374754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,4,64,power_law_1.01,0.04043136119842529
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,4,64,power_law_1.01,0.04303232192993164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,4,64,power_law_1.01,0.04318079948425293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,4,64,power_law_1.01,0.045411839485168456
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,4,64,power_law_1.01,0.050831360816955565
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,4,64,power_law_1.01,0.051823358535766605
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,4,64,power_law_1.01,0.06033408164978027
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,4,64,power_law_1.01,0.0728934383392334
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,4,64,power_law_1.01,0.10653311729431152
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,4,64,power_law_1.01,0.12046463966369629
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,4,64,power_law_1.01,0.15179648399353027
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,4,64,power_law_1.01,0.23952255249023438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,4,64,power_law_1.2,0.020341761112213134
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,4,64,power_law_1.2,0.020776960849761963
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,4,64,power_law_1.01,0.034647040367126465
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,4,64,power_law_1.01,0.3650624084472656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,4,64,power_law_1.01,0.03353984117507934
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,4,64,power_law_1.2,0.02327552080154419
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,4,64,power_law_1.01,0.45985153198242185
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,4,64,power_law_1.01,0.03865472078323364
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,4,64,power_law_1.2,0.033365759849548336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,4,64,power_law_1.2,0.030585598945617676
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,4,64,power_law_1.2,0.03797375917434692
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,4,64,power_law_1.2,0.04466303825378418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,4,64,power_law_1.2,0.03544447898864746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,4,64,power_law_1.2,0.046289920806884766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,4,64,power_law_1.2,0.05318528175354004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,4,64,power_law_1.2,0.05465472221374512
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,4,64,power_law_1.2,0.060682239532470705
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,4,64,power_law_1.2,0.06219136238098144
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,4,64,power_law_1.2,0.07194111824035646
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,4,64,power_law_1.2,0.07974783897399902
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,4,64,power_law_1.2,0.10384511947631836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,4,64,power_law_1.2,0.11668992042541504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,4,64,power_law_1.2,0.020986878871917726
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,4,64,power_law_1.2,0.17773311614990234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,4,64,power_law_1.2,0.02962559938430786
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,4,64,power_law_1.2,0.2787558364868164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,4,64,power_law_1.2,0.048623361587524415
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,4,64,power_law_1.2,0.02225152015686035
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,4,64,power_law_1.2,0.4272512054443359
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,4,64,power_law_1.2,0.06050303936004638
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,4,64,power_law_1.2,0.022497279644012453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,4,64,power_law_1.2,0.5476083374023437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,4,64,power_law_1.2,0.022348799705505372
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,4,64,power_law_1.2,0.022571520805358885
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,4,64,power_law_1.2,0.03405951976776123
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,4,64,power_law_1.2,0.035088639259338375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,4,64,power_law_1.2,0.0348198390007019
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,4,64,power_law_1.2,0.035518720149993896
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,4,64,power_law_1.2,0.035767040252685546
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,4,64,power_law_1.2,0.03622015953063965
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,4,64,power_law_1.2,0.037639679908752444
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,4,64,power_law_1.2,0.20669824600219727
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,4,64,power_law_1.2,0.039944961071014404
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,4,64,power_law_1.2,0.043727359771728515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,4,64,power_law_1.2,0.04333439826965332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,4,64,power_law_1.2,0.04652544021606445
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,4,64,power_law_1.2,0.05281023979187012
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,4,64,power_law_1.2,0.057414398193359376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,4,64,power_law_1.2,0.0756057596206665
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,4,64,power_law_1.2,0.09048576354980468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,4,64,power_law_1.2,0.12483200073242187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,4,64,power_law_1.2,0.14833536148071289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,4,64,power_law_1.2,0.19821952819824218
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,4,64,power_law_1.2,0.8364134216308594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,8,32,balanced,0.019288320541381836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,4,64,power_law_1.2,0.3188595199584961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,8,32,balanced,0.019141119718551636
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,8,32,balanced,0.019463679790496825
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,8,32,balanced,0.018836480379104615
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,4,64,power_law_1.2,0.03879679918289185
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,8,32,balanced,0.02025984048843384
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,8,32,balanced,0.03340159893035889
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,8,32,balanced,0.032590079307556155
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,8,32,balanced,0.032620799541473386
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,8,32,balanced,0.036528639793395996
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,4,64,power_law_1.2,0.6329779052734376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,8,32,balanced,0.039256319999694825
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,8,32,balanced,0.04107391834259033
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,8,32,balanced,0.03943295955657959
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,8,32,balanced,0.043784961700439454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,8,32,balanced,0.056907520294189454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,8,32,balanced,0.05202559947967529
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,8,32,balanced,0.056017918586730955
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,8,32,balanced,0.062001919746398924
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,8,32,balanced,0.06775680065155029
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,8,32,balanced,0.07333119869232177
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,8,32,balanced,0.08635775566101074
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,8,32,balanced,0.09841919898986816
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,8,32,balanced,0.12418687820434571
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,8,32,balanced,0.14823424339294433
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,8,32,balanced,0.19769983291625975
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,8,32,balanced,0.021736960411071777
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,8,32,balanced,0.021848320960998535
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,8,32,balanced,0.2446553611755371
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,8,32,balanced,0.02166271924972534
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,8,32,balanced,0.0220032000541687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,4,64,power_law_1.2,0.042764801979064945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,8,32,balanced,0.0222489595413208
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,8,32,balanced,0.02473599910736084
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,8,32,balanced,0.034883840084075926
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,8,32,balanced,0.0349452805519104
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,8,32,balanced,0.03260288000106811
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,8,32,balanced,0.03584000110626221
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,8,32,balanced,0.03581056118011475
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,8,32,balanced,0.03668864011764526
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,8,32,balanced,0.039202558994293216
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,8,32,balanced,0.037304320335388184
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,8,32,balanced,0.04025728225708008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,8,32,balanced,0.04128511905670166
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,8,32,balanced,0.0427788782119751
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,8,32,balanced,0.03863679885864258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,8,32,balanced,0.043884801864624026
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,8,32,balanced,0.04715007781982422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,8,32,balanced,0.052080640792846675
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,8,32,balanced,0.05617792129516601
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,8,32,balanced,0.06912896156311035
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,8,32,balanced,0.0820736026763916
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,8,32,balanced,0.10879743576049805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,8,32,balanced,0.1308249568939209
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,8,32,balanced,0.17992704391479492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,8,32,power_law_1.01,0.021172480583190916
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,8,32,balanced,0.22489471435546876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,8,32,power_law_1.01,0.02135551929473877
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,8,32,power_law_1.01,0.021575679779052732
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,8,32,power_law_1.01,0.022332160472869872
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,8,32,balanced,0.022352640628814695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,8,32,power_law_1.01,0.03489023923873901
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,8,32,power_law_1.01,0.03607039928436279
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,8,32,power_law_1.01,0.03653120040893555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,8,32,power_law_1.01,0.046927361488342284
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,8,32,power_law_1.01,0.03835520029067993
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,8,32,power_law_1.01,0.03996543884277344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,8,32,power_law_1.01,0.04194303989410401
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,8,32,power_law_1.01,0.05051392078399658
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,8,32,power_law_1.01,0.054978561401367185
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,8,32,power_law_1.01,0.0599616003036499
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,8,32,power_law_1.01,0.05382783889770508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,8,32,power_law_1.01,0.05374720096588135
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,8,32,power_law_1.01,0.05951231956481934
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,8,32,power_law_1.01,0.06946559906005859
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,8,32,power_law_1.01,0.07227647781372071
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,8,32,power_law_1.01,0.08485504150390624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,8,32,power_law_1.01,0.09753984451293946
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,8,32,power_law_1.01,0.12642815589904785
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,8,32,power_law_1.01,0.16457088470458986
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,8,32,power_law_1.01,0.22188159942626956
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,8,32,power_law_1.01,0.2967296028137207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,8,32,power_law_1.01,0.02351360082626343
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,8,32,power_law_1.01,0.023562240600585937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,4,64,power_law_1.2,0.4649548721313477
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,8,32,power_law_1.01,0.4218585586547851
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,8,32,power_law_1.01,0.02376703977584839
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,8,32,power_law_1.01,0.02413055896759033
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,8,32,power_law_1.01,0.03448832035064697
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,8,32,power_law_1.01,0.5554636764526367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,8,32,power_law_1.01,0.03529855966567993
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,8,32,power_law_1.01,0.03661695957183838
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,8,32,power_law_1.01,0.03687808036804199
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,8,32,power_law_1.01,0.038204159736633304
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,8,32,power_law_1.01,0.039959039688110355
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,8,32,power_law_1.01,0.041335039138793946
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,8,32,power_law_1.01,0.042769918441772456
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,8,32,power_law_1.01,0.04078080177307129
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,8,32,power_law_1.01,0.04438784122467041
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,8,32,power_law_1.01,0.04798336029052734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,8,32,power_law_1.01,0.05375999927520751
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,8,32,power_law_1.01,0.055749120712280276
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,8,32,power_law_1.01,0.0721062421798706
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,8,32,power_law_1.01,0.08450943946838378
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,8,32,power_law_1.01,0.10425344467163086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,8,32,power_law_1.01,0.15199104309082032
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,8,32,power_law_1.01,0.19032575607299804
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,8,32,power_law_1.01,0.26002687454223633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,8,32,power_law_1.2,0.021227519512176513
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,8,32,power_law_1.01,0.3739110565185547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,8,32,power_law_1.2,0.021459200382232667
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,8,32,power_law_1.2,0.02149120092391968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,8,32,power_law_1.01,0.40040576934814454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,8,32,power_law_1.2,0.02144511938095093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,8,32,power_law_1.2,0.030336000919342042
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,8,32,power_law_1.2,0.03349119901657104
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,8,32,power_law_1.2,0.03727488040924072
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,8,32,power_law_1.2,0.035916800498962405
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,8,32,power_law_1.2,0.04094207763671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,8,32,power_law_1.2,0.05056128025054931
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,8,32,power_law_1.2,0.052030720710754395
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,8,32,power_law_1.2,0.05629695892333984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,8,32,power_law_1.2,0.05800447940826416
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,8,32,power_law_1.2,0.05697792053222657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,8,32,power_law_1.2,0.0643174409866333
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,8,32,power_law_1.2,0.07116032123565674
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,8,32,power_law_1.2,0.07763199806213379
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,8,32,power_law_1.2,0.09838720321655273
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,8,32,power_law_1.01,0.035073280334472656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,8,32,power_law_1.2,0.12116607666015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,8,32,power_law_1.01,0.037918720245361325
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,8,32,power_law_1.2,0.15661439895629883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,8,32,power_law_1.2,0.19156864166259766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,8,32,power_law_1.2,0.023417599201202392
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,8,32,power_law_1.2,0.030074880123138425
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,8,32,power_law_1.2,0.3149260711669922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,8,32,power_law_1.2,0.038926079273223876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,8,32,power_law_1.2,0.3435353469848633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,8,32,power_law_1.2,0.023713281154632566
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,8,32,power_law_1.2,0.023875839710235595
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,8,32,power_law_1.2,0.023841280937194825
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,8,32,power_law_1.2,0.5261798477172851
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,8,32,power_law_1.2,0.03439359903335572
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,8,32,power_law_1.2,0.035048959255218504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,8,32,power_law_1.2,0.03564671993255615
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,8,32,power_law_1.2,0.03689984083175659
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,8,32,power_law_1.2,0.036550400257110593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,8,32,power_law_1.2,0.7086656188964844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,8,32,power_law_1.2,0.037119998931884765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,8,32,power_law_1.2,0.03854847908020019
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,8,32,power_law_1.2,0.04112127780914306
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,8,32,power_law_1.2,0.04365824222564697
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,8,32,power_law_1.2,0.043162879943847654
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,8,32,power_law_1.2,0.045583357810974126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,8,32,power_law_1.2,0.050000638961791986
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,8,32,power_law_1.2,0.0538265609741211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,8,32,power_law_1.2,0.061941761970520016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,8,32,power_law_1.2,0.08073984146118164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,8,32,power_law_1.2,0.10313599586486817
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,8,32,power_law_1.2,0.14215423583984374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,8,32,power_law_1.2,0.1667545509338379
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,16,16,balanced,0.019224319458007812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,16,16,balanced,0.019353599548339845
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,8,32,power_law_1.2,0.30787967681884765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,16,16,balanced,0.020071680545806887
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,16,16,balanced,0.020697600841522217
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,8,32,power_law_1.2,0.45214721679687503
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,16,16,balanced,0.029707520008087157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,16,16,balanced,0.034115839004516604
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,16,16,balanced,0.0489792013168335
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,16,16,balanced,0.03770240068435669
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,8,32,power_law_1.2,0.6294553756713868
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,8,32,power_law_1.2,0.04233215808868408
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,16,16,balanced,0.05025279998779297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,16,16,balanced,0.05456639766693115
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,16,16,balanced,0.05754752159118652
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,16,16,balanced,0.051696639060974124
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,16,16,balanced,0.05499392032623292
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,16,16,balanced,0.05669119834899903
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,16,16,balanced,0.06066944122314453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,16,16,balanced,0.06678912162780762
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,16,16,balanced,0.07165823936462402
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,16,16,balanced,0.08426624298095703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,16,16,balanced,0.09624704360961914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,8,32,power_law_1.2,0.22305791854858398
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,16,16,balanced,0.12125696182250975
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,16,16,balanced,0.14302080154418945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,16,16,balanced,0.18976640701293945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,16,16,balanced,0.021857280731201172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,16,16,balanced,0.02219520092010498
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,16,16,balanced,0.022023680210113524
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,16,16,balanced,0.3256025695800781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,16,16,balanced,0.02230655908584595
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,16,16,balanced,0.413702392578125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,16,16,balanced,0.022557439804077147
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,16,16,balanced,0.047425279617309565
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,16,16,balanced,0.024067840576171874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,16,16,balanced,0.02893440008163452
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,16,16,balanced,0.042109441757202146
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,16,16,balanced,0.04115839958190918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,16,16,balanced,0.04172416210174561
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,16,16,balanced,0.04288640022277832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,16,16,balanced,0.0445363187789917
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,16,16,balanced,0.0456063985824585
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,16,16,balanced,0.04223487854003906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,16,16,balanced,0.045423359870910646
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,16,16,balanced,0.04570752143859863
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,16,16,balanced,0.04882944107055664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,16,16,balanced,0.053326082229614255
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,16,16,balanced,0.059123201370239256
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,16,16,balanced,0.07050752162933349
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,16,16,balanced,0.08130816459655762
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,16,16,balanced,0.10542847633361815
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,16,16,balanced,0.12666624069213867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,16,16,balanced,0.17666303634643554
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,16,16,balanced,0.21614080429077148
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,16,16,balanced,0.23406591415405273
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,16,16,power_law_1.01,0.021199359893798828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,16,16,balanced,0.3098380851745605
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,16,16,power_law_1.01,0.027000319957733155
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,16,16,power_law_1.01,0.02695296049118042
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,16,16,power_law_1.01,0.02970495939254761
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,16,16,power_law_1.01,0.0308953595161438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,16,16,power_law_1.01,0.04404607772827148
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,16,16,power_law_1.01,0.042018561363220214
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,16,16,power_law_1.01,0.048162560462951656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,16,16,power_law_1.01,0.05349247932434083
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,16,16,power_law_1.01,0.05296895980834961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,16,16,power_law_1.01,0.05626368045806884
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,16,16,power_law_1.01,0.05666304111480713
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,16,16,power_law_1.01,0.05589248180389404
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,16,16,power_law_1.01,0.058209280967712405
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,16,16,power_law_1.01,0.06439551830291748
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,16,16,power_law_1.01,0.07216512203216553
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,16,16,power_law_1.01,0.08484224319458009
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,16,16,power_law_1.01,0.09808128356933594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,16,16,power_law_1.01,0.11505791664123535
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,16,16,power_law_1.01,0.14550016403198243
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,16,16,power_law_1.01,0.18288896560668946
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,16,16,balanced,0.39950080871582033
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,16,16,power_law_1.01,0.24905216217041018
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,16,16,power_law_1.01,0.04384640216827392
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,16,16,power_law_1.01,0.3067814445495606
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,16,16,power_law_1.01,0.02319999933242798
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,16,16,power_law_1.01,0.025968639850616454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,16,16,power_law_1.01,0.05881599903106689
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,16,16,power_law_1.01,0.4586316680908203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,16,16,power_law_1.01,0.02599551916122437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,16,16,power_law_1.01,0.026812160015106203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,16,16,power_law_1.01,0.02813184022903442
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,16,16,power_law_1.01,0.547504653930664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,16,16,power_law_1.01,0.03807104110717773
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,16,16,power_law_1.01,0.04111999988555908
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,16,16,power_law_1.01,0.04131968021392822
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,16,16,power_law_1.01,0.04226687908172607
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,16,16,power_law_1.01,0.044038400650024415
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,16,16,power_law_1.01,0.04501376152038574
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,16,16,power_law_1.01,0.04535295963287354
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,16,16,power_law_1.01,0.043308801651000976
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,16,16,power_law_1.01,0.046384639739990234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,16,16,power_law_1.01,0.04822144031524658
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,16,16,power_law_1.01,0.06770175933837891
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,16,16,power_law_1.01,0.08663935661315918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,16,16,power_law_1.01,0.09635199546813965
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,16,16,power_law_1.01,0.13283072471618654
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,16,16,power_law_1.01,0.17794303894042968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,16,16,power_law_1.01,0.24049152374267577
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,16,16,power_law_1.01,0.040110077857971195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,16,16,power_law_1.01,0.2926988792419434
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,16,16,power_law_1.2,0.0210534405708313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,16,16,power_law_1.01,0.4174886322021484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,16,16,power_law_1.01,0.051824641227722165
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,16,16,power_law_1.2,0.02600575923919678
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,16,16,power_law_1.01,0.05957888126373291
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,16,16,balanced,0.041547517776489254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,16,16,power_law_1.2,0.028729600906372072
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,16,16,power_law_1.2,0.03935744047164917
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,16,16,power_law_1.2,0.03704832077026367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,16,16,power_law_1.2,0.048419837951660154
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,16,16,power_law_1.2,0.040215039253234865
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,16,16,power_law_1.2,0.04826367855072021
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,16,16,power_law_1.2,0.05489664077758789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,16,16,power_law_1.2,0.05609600067138672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,16,16,power_law_1.2,0.056167678833007814
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,16,16,power_law_1.2,0.05502975940704345
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,16,16,power_law_1.2,0.05694975852966309
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,16,16,power_law_1.2,0.061098241806030275
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,16,16,power_law_1.2,0.06564735889434814
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,16,16,power_law_1.2,0.07603839874267579
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,16,16,power_law_1.2,0.08774527549743652
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,16,16,power_law_1.2,0.09844479560852051
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,16,16,power_law_1.2,0.12310400009155273
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,16,16,power_law_1.01,0.5006156921386719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,16,16,power_law_1.2,0.1661631965637207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,16,16,power_law_1.2,0.18593408584594726
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,16,16,power_law_1.2,0.2801228713989258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,16,16,power_law_1.2,0.35189376831054686
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,16,16,power_law_1.2,0.023398399353027344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,16,16,power_law_1.2,0.026314239501953125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,16,16,power_law_1.2,0.02609407901763916
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,16,16,power_law_1.2,0.5669196701049805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,16,16,power_law_1.2,0.026131200790405273
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,16,16,power_law_1.2,0.03927167892456055
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,16,16,power_law_1.2,0.04071040153503418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,16,16,power_law_1.2,0.040864000320434565
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,16,16,power_law_1.2,0.04123392105102539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,16,16,power_law_1.2,0.027691519260406493
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,16,16,power_law_1.2,0.02839168071746826
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,16,16,power_law_1.2,0.04407167911529541
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,16,16,power_law_1.2,0.04623104095458984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,16,16,power_law_1.2,0.04403711795806885
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,16,16,power_law_1.2,0.04470399856567383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,16,16,power_law_1.2,0.046242561340332035
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,16,16,power_law_1.2,0.04882944107055664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,16,16,power_law_1.2,0.05386623859405517
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,16,16,power_law_1.2,0.0636300802230835
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,16,16,power_law_1.2,0.07418623924255371
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,16,16,power_law_1.2,0.08801024436950684
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,16,16,power_law_1.2,0.10931967735290526
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,16,16,power_law_1.2,0.15807616233825683
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,16,16,power_law_1.2,0.026645760536193847
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,16,16,power_law_1.2,0.19094272613525393
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,16,16,power_law_1.2,0.2516684722900391
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,32,8,balanced,0.019760639667510987
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,32,8,balanced,0.020490241050720216
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,16,16,power_law_1.2,0.8053401947021485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,16,16,power_law_1.2,0.3186291122436523
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,32,8,balanced,0.023168001174926758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,32,8,balanced,0.026254079341888427
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,16,16,power_law_1.2,0.04249728202819824
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,32,8,balanced,0.03659775972366333
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,32,8,balanced,0.051774721145629886
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,32,8,balanced,0.057644801139831545
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,32,8,balanced,0.06099967956542969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,32,8,balanced,0.053107199668884275
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,32,8,balanced,0.056897277832031246
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,32,8,balanced,0.05833343982696533
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,32,8,balanced,0.06317823886871338
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,32,8,balanced,0.06584832191467285
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,32,8,balanced,0.06875264167785644
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,32,8,balanced,0.07385983943939209
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,32,8,balanced,0.08547967910766602
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,32,8,balanced,0.09684608459472657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,32,8,balanced,0.11974656105041503
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,32,8,balanced,0.1406886386871338
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,32,8,balanced,0.18650367736816406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,32,8,balanced,0.22983167648315428
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,32,8,balanced,0.31722240447998046
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,32,8,balanced,0.021639680862426756
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,32,8,balanced,0.40350463867187497
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,32,8,balanced,0.022547199726104736
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,32,8,balanced,0.023093760013580322
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,32,8,balanced,0.023517439365386965
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,16,16,power_law_1.2,0.5372966384887695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,32,8,balanced,0.02492160081863403
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,16,16,power_law_1.2,0.5880384063720703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,32,8,balanced,0.5841100692749024
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,32,8,balanced,0.035865600109100344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,32,8,balanced,0.050690560340881353
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,32,8,balanced,0.05210879802703857
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,32,8,balanced,0.7710873413085937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,32,8,balanced,0.05003647804260254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,32,8,balanced,0.05106688022613526
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,32,8,balanced,0.05612287998199463
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,32,8,balanced,0.05726208209991455
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,32,8,balanced,0.06047359943389893
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,32,8,balanced,0.06452991962432861
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,32,8,balanced,0.07730048179626466
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,32,8,balanced,0.05415040016174316
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,32,8,balanced,0.08840703964233398
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,32,8,balanced,0.1115993595123291
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,32,8,balanced,0.028509440422058108
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,32,8,balanced,0.17558912277221678
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,32,8,balanced,0.049314560890197756
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,32,8,balanced,0.21437824249267576
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,32,8,balanced,0.053300480842590335
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,32,8,balanced,0.3139814376831055
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,32,8,balanced,0.053840641975402836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,32,8,balanced,0.3944755172729492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,32,8,power_law_1.01,0.02178431987762451
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,32,8,power_law_1.01,0.023585278987884522
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,32,8,balanced,0.5744230270385742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,32,8,power_law_1.01,0.025004799365997317
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,32,8,power_law_1.01,0.034378240108489985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,32,8,power_law_1.01,0.034579200744628905
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,32,8,balanced,0.7564736175537109
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,32,8,power_law_1.01,0.05365375995635986
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,32,8,balanced,0.13237504005432127
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,32,8,power_law_1.01,0.05622655868530273
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,32,8,power_law_1.01,0.05796351909637452
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,32,8,power_law_1.01,0.06376192092895508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,32,8,power_law_1.01,0.05527935981750488
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,32,8,power_law_1.01,0.058003201484680175
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,32,8,power_law_1.01,0.06238848209381104
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,32,8,power_law_1.01,0.07654272079467773
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,32,8,power_law_1.01,0.09144960403442383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,32,8,power_law_1.01,0.10862208366394044
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,32,8,power_law_1.01,0.13107839584350586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,32,8,power_law_1.01,0.030209279060363768
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,32,8,power_law_1.01,0.18000255584716798
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,32,8,power_law_1.01,0.20624000549316407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,32,8,power_law_1.01,0.052872958183288576
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,32,8,power_law_1.01,0.05215616226196289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,32,8,power_law_1.01,0.2822220802307129
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,32,8,power_law_1.01,0.024084479808807374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,32,8,power_law_1.01,0.3851532745361328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,32,8,power_law_1.01,0.02681727886199951
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,32,8,power_law_1.01,0.028977921009063722
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,32,8,power_law_1.01,0.06763775825500488
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,32,8,power_law_1.01,0.030030078887939453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,32,8,power_law_1.01,0.0318937611579895
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,32,8,power_law_1.01,0.04747776031494141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,32,8,power_law_1.01,0.04825471878051758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,32,8,power_law_1.01,0.7164173126220703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,32,8,power_law_1.01,0.04853631973266602
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,32,8,power_law_1.01,0.049928960800170896
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,32,8,power_law_1.01,0.053096961975097653
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,32,8,power_law_1.01,0.050481920242309575
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,32,8,power_law_1.01,0.05294335842132568
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,32,8,power_law_1.01,0.05029248237609864
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,32,8,power_law_1.01,0.05538432121276855
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,32,8,power_law_1.01,0.05616896152496338
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,32,8,power_law_1.01,0.059006719589233404
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,32,8,power_law_1.01,0.06871039867401123
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,32,8,power_law_1.01,0.08291711807250976
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,32,8,power_law_1.01,0.10348416328430177
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,32,8,power_law_1.01,0.12302463531494141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,32,8,power_law_1.01,0.16611967086791993
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,32,8,power_law_1.01,0.06415232181549071
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,32,8,power_law_1.01,0.19720703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,32,8,power_law_1.01,0.0334553599357605
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,32,8,power_law_1.01,0.2786278343200684
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,32,8,power_law_1.2,0.02165503978729248
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,32,8,power_law_1.2,0.023505918979644776
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,32,8,power_law_1.2,0.023738880157470704
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,32,8,power_law_1.2,0.027066879272460938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,32,8,power_law_1.01,0.5093171310424804
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,32,8,power_law_1.2,0.03401472091674805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,32,8,power_law_1.2,0.0523737621307373
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,32,8,power_law_1.2,0.0373964810371399
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,32,8,power_law_1.2,0.04986752033233642
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,32,8,power_law_1.01,0.6703705596923828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,32,8,power_law_1.2,0.06328320026397705
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,32,8,power_law_1.2,0.059415040016174314
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,32,8,power_law_1.2,0.05470208168029785
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,32,8,power_law_1.2,0.05955455780029297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,32,8,power_law_1.2,0.06152575969696045
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,32,8,power_law_1.2,0.06545919895172118
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,32,8,power_law_1.2,0.07072127819061279
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,32,8,power_law_1.01,0.5439257431030273
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,32,8,power_law_1.2,0.0822822380065918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,32,8,power_law_1.2,0.09346559524536133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,32,8,power_law_1.2,0.11154687881469727
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,32,8,power_law_1.01,0.37133056640625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,32,8,power_law_1.2,0.13516544342041015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,32,8,power_law_1.2,0.179420166015625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,32,8,power_law_1.2,0.2238937568664551
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,32,8,power_law_1.2,0.02421247959136963
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,32,8,power_law_1.2,0.40202495574951175
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,32,8,power_law_1.2,0.051951360702514646
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,32,8,power_law_1.2,0.026931200027465817
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,32,8,power_law_1.2,0.5765798568725586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,32,8,power_law_1.2,0.027860479354858397
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,32,8,power_law_1.2,0.029072639942169187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,32,8,power_law_1.2,0.8063346862792968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,32,8,power_law_1.2,0.031110401153564456
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,32,8,power_law_1.2,0.0335423994064331
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,32,8,power_law_1.2,0.04839935779571533
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,32,8,power_law_1.2,0.049434881210327145
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,32,8,power_law_1.2,0.049989118576049804
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,32,8,power_law_1.2,0.05133567810058594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,32,8,power_law_1.2,0.04915200233459473
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,32,8,power_law_1.2,0.051441922187805175
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,32,8,power_law_1.2,0.05511936187744141
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,32,8,power_law_1.2,0.05586431980133057
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,32,8,power_law_1.2,0.060862717628479
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,32,8,power_law_1.2,0.07400320053100587
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,32,8,power_law_1.2,0.08560383796691895
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,32,8,power_law_1.2,0.10512255668640136
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,32,8,power_law_1.2,0.13150079727172853
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,32,8,power_law_1.2,0.048222718238830564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,32,8,power_law_1.2,0.1722559928894043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,32,8,power_law_1.2,0.21165184020996092
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,32,8,power_law_1.2,0.05082111835479737
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,32,8,power_law_1.2,0.3192998313903809
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,32,8,power_law_1.2,0.3317273712158203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,32,8,power_law_1.2,0.38906623840332033
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,2,128,balanced,0.020047359466552735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,32,8,power_law_1.2,0.05707263946533203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,2,128,balanced,0.020044798851013186
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,2,128,balanced,0.019992320537567138
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,2,128,balanced,0.020042240619659424
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,2,128,balanced,0.028161280155181885
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,2,128,balanced,0.028441600799560547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,2,128,balanced,0.02833280086517334
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,2,128,balanced,0.02847232103347778
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,2,128,balanced,0.028698880672454834
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,2,128,balanced,0.028671998977661133
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,2,128,balanced,0.028980479240417478
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,2,128,balanced,0.02929663896560669
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,32,8,power_law_1.2,0.710167007446289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,2,128,balanced,0.030700800418853758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,2,128,balanced,0.03216639995574951
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,2,128,balanced,0.03172096014022827
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,2,128,balanced,0.03432575941085815
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,2,128,balanced,0.03957632064819336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,2,128,balanced,0.034647040367126465
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,2,128,balanced,0.04590335845947265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,2,128,balanced,0.042789120674133305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,2,128,balanced,0.05453824043273926
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,2,128,balanced,0.06357376098632812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,2,128,balanced,0.07778560161590577
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,2,128,balanced,0.08766464233398438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,2,128,balanced,0.01993600010871887
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,2,128,power_law_1.01,0.019834879636764526
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,2,128,power_law_1.01,0.019775999784469606
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,2,128,power_law_1.01,0.0199616003036499
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,2,128,power_law_1.01,0.028158719539642336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,2,128,power_law_1.01,0.02832895994186401
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,2,128,power_law_1.01,0.028550400733947757
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,2,128,power_law_1.01,0.029295361042022704
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,2,128,power_law_1.01,0.030570240020751954
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,2,128,power_law_1.01,0.031247360706329348
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,2,128,power_law_1.01,0.03195264101028442
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,2,128,power_law_1.01,0.03341056108474731
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,2,128,power_law_1.01,0.035306239128112794
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,2,128,power_law_1.01,0.0377676796913147
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,2,128,power_law_1.01,0.039851520061492916
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,2,128,power_law_1.01,0.040885758399963376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,2,128,balanced,0.03540607929229737
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,2,128,power_law_1.01,0.04954368114471436
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,2,128,balanced,0.03949951887130737
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,2,128,power_law_1.01,0.0565555191040039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,2,128,power_law_1.01,0.061919999122619626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,2,128,power_law_1.01,0.0761075210571289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,2,128,power_law_1.01,0.0867686367034912
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,2,128,power_law_1.01,0.12788352012634277
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,2,128,power_law_1.01,0.14605440139770506
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,2,128,power_law_1.01,0.21160703659057617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,2,128,power_law_1.2,0.019924479722976684
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,2,128,power_law_1.2,0.019704320430755616
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,2,128,power_law_1.2,0.02804608106613159
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,32,8,power_law_1.2,0.5756121444702148
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,2,128,power_law_1.2,0.02829567909240723
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,2,128,power_law_1.2,0.029122560024261473
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,2,128,power_law_1.2,0.02880000114440918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,2,128,power_law_1.01,0.633070068359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,2,128,power_law_1.2,0.029605119228363036
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,2,128,power_law_1.2,0.03347455978393555
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,2,128,power_law_1.2,0.0333350396156311
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,2,128,power_law_1.2,0.03542272090911865
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,2,128,power_law_1.2,0.03720191955566406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,2,128,power_law_1.2,0.04007423877716064
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,2,128,power_law_1.2,0.04599296092987061
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,2,128,power_law_1.2,0.04770688056945801
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,2,128,power_law_1.2,0.05464960098266601
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,2,128,power_law_1.2,0.060834560394287106
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,2,128,power_law_1.01,0.29243520736694334
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,2,128,power_law_1.2,0.09469568252563476
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,2,128,power_law_1.01,0.46372222900390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,2,128,power_law_1.2,0.13035391807556151
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,2,128,power_law_1.01,0.03573632001876831
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,2,128,power_law_1.2,0.19152256011962893
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,2,128,power_law_1.2,0.0311244797706604
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,2,128,power_law_1.2,0.24685440063476563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,2,128,power_law_1.2,0.03547264099121093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,4,64,balanced,0.018539520502090453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,4,64,balanced,0.01835904002189636
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,4,64,balanced,0.018611199855804443
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,4,64,balanced,0.019281920194625855
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,2,128,power_law_1.2,0.0765337610244751
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,4,64,balanced,0.02661504030227661
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,4,64,balanced,0.026860799789428715
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,4,64,balanced,0.027086079120635986
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,4,64,balanced,0.02743168115615845
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,2,128,power_law_1.2,0.4258700942993164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,4,64,balanced,0.027331840991973878
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,4,64,balanced,0.028120319843292236
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,4,64,balanced,0.027974400520324703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,4,64,balanced,0.02857856035232544
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,4,64,balanced,0.029655039310455322
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,4,64,balanced,0.029687039852142334
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,4,64,balanced,0.031816959381103516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,4,64,balanced,0.03169663906097412
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,4,64,balanced,0.03372159957885742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,4,64,balanced,0.035665919780731206
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,4,64,balanced,0.04111231803894043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,4,64,balanced,0.0427507209777832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,4,64,balanced,0.050690560340881353
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,4,64,balanced,0.05670911788940429
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,4,64,balanced,0.06813695907592773
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,4,64,balanced,0.07782527923583984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,4,64,balanced,0.095797758102417
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,4,64,balanced,0.11268351554870606
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,4,64,balanced,0.02744191884994507
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,4,64,power_law_1.01,0.0188428795337677
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,4,64,power_law_1.01,0.019381760358810424
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,4,64,power_law_1.01,0.018776320219039917
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,4,64,power_law_1.01,0.02671744108200073
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,4,64,power_law_1.01,0.026856958866119385
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,4,64,power_law_1.01,0.027488000392913818
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,4,64,power_law_1.01,0.027784960269927977
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,4,64,power_law_1.01,0.02891263961791992
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,4,64,power_law_1.01,0.02875648021697998
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,4,64,power_law_1.01,0.02986112117767334
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,4,64,power_law_1.01,0.03159296035766602
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,4,64,power_law_1.01,0.0326963210105896
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,4,64,power_law_1.01,0.03522559881210327
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,4,64,power_law_1.01,0.037719039916992186
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,4,64,power_law_1.01,0.040604162216186526
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,4,64,power_law_1.01,0.05041279792785645
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,4,64,power_law_1.01,0.05202816009521485
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,4,64,power_law_1.01,0.059658241271972653
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,4,64,power_law_1.01,0.07583231925964355
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,4,64,power_law_1.01,0.0805337619781494
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,4,64,power_law_1.01,0.10627327919006349
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,4,64,power_law_1.01,0.14176511764526367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,2,128,power_law_1.2,0.5617446517944336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,4,64,power_law_1.01,0.2043903923034668
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,4,64,power_law_1.01,0.23662591934204102
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,4,64,power_law_1.2,0.018826240301132204
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,4,64,power_law_1.2,0.019456000328063966
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,4,64,power_law_1.01,0.36163585662841796
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,4,64,power_law_1.2,0.019144959449768066
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,4,64,power_law_1.2,0.02676352024078369
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,4,64,power_law_1.2,0.0272652792930603
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,4,64,power_law_1.2,0.027248640060424805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,4,64,power_law_1.2,0.028446719646453855
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,4,64,power_law_1.2,0.028390400409698487
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,4,64,power_law_1.2,0.02964479923248291
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,4,64,power_law_1.2,0.032522239685058595
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,4,64,power_law_1.2,0.03295743942260742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,4,64,power_law_1.2,0.03568639993667603
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,4,64,power_law_1.2,0.0379801607131958
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,4,64,power_law_1.2,0.04478464126586914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,4,64,power_law_1.2,0.04841343879699707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,4,64,power_law_1.2,0.04998400211334229
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,2,128,power_law_1.2,0.8462489318847656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,4,64,power_law_1.2,0.057239041328430176
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,4,64,power_law_1.2,0.07008255958557129
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,4,64,power_law_1.2,0.07938432216644287
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,4,64,power_law_1.01,0.03191423892974853
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,4,64,power_law_1.01,0.4770393753051758
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,4,64,power_law_1.2,0.19453439712524415
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,4,64,power_law_1.2,0.03024768114089966
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,8,32,balanced,0.01968000054359436
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,8,32,balanced,0.019687680006027223
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,8,32,balanced,0.02002432107925415
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,4,64,power_law_1.2,0.32104705810546874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,8,32,balanced,0.0212774395942688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,4,64,power_law_1.2,0.09810815811157227
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,8,32,balanced,0.030389759540557858
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,4,64,power_law_1.2,0.155918083190918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,8,32,balanced,0.0329804801940918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,8,32,balanced,0.03292927980422974
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,8,32,balanced,0.033145599365234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,8,32,balanced,0.03369215965270996
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,8,32,balanced,0.032076799869537355
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,8,32,balanced,0.03263999938964844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,8,32,balanced,0.03524352073669433
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,8,32,balanced,0.034190080165863036
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,4,64,power_law_1.2,0.3871116638183594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,8,32,balanced,0.03745536088943481
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,8,32,balanced,0.038609919548034666
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,8,32,balanced,0.0404748821258545
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,8,32,balanced,0.04540287971496582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,8,32,balanced,0.050623998641967774
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,8,32,balanced,0.05470592021942139
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,8,32,balanced,0.057480959892272955
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,8,32,balanced,0.07466879844665528
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,8,32,balanced,0.08348416328430176
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,8,32,balanced,0.10278783798217775
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,8,32,balanced,0.11889792442321778
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,8,32,balanced,0.15622783660888673
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,8,32,balanced,0.20102783203125002
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,8,32,power_law_1.01,0.02142848014831543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,8,32,power_law_1.01,0.021518080234527587
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,8,32,power_law_1.01,0.02209280014038086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,8,32,power_law_1.01,0.030164480209350586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,8,32,power_law_1.01,0.03074559926986694
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,8,32,power_law_1.01,0.03139584064483643
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,8,32,power_law_1.01,0.033518080711364744
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,8,32,power_law_1.01,0.0331712007522583
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,8,32,power_law_1.01,0.03431551933288574
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,8,32,balanced,0.03014656066894531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,8,32,power_law_1.01,0.03517823934555054
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,8,32,power_law_1.01,0.036661760807037355
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,8,32,power_law_1.01,0.03879168033599854
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,8,32,power_law_1.01,0.04036096096038818
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,8,32,power_law_1.01,0.04276735782623291
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,8,32,power_law_1.01,0.04741631984710694
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,8,32,power_law_1.01,0.05312640190124511
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,8,32,power_law_1.01,0.0559884786605835
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,8,32,power_law_1.01,0.06542975902557373
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,8,32,power_law_1.01,0.07541632175445556
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,8,32,power_law_1.01,0.09152511596679688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,8,32,power_law_1.01,0.10926207542419433
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,8,32,power_law_1.01,0.14934271812438965
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,8,32,power_law_1.01,0.18342144012451173
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,8,32,power_law_1.01,0.25881984710693356
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,8,32,power_law_1.2,0.021416959762573243
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,8,32,power_law_1.01,0.35067775726318356
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,8,32,power_law_1.2,0.0217523193359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,8,32,power_law_1.01,0.4901772689819336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,8,32,power_law_1.2,0.02189055919647217
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,8,32,power_law_1.2,0.029855999946594235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,4,64,power_law_1.2,0.645006103515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,8,32,power_law_1.2,0.030507519245147708
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,8,32,power_law_1.2,0.03221247911453247
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,8,32,power_law_1.2,0.03258624076843262
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,8,32,power_law_1.2,0.03327359914779663
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,8,32,power_law_1.2,0.0349401593208313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,8,32,power_law_1.2,0.036735999584198
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,8,32,power_law_1.2,0.0405350399017334
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,8,32,power_law_1.2,0.046558718681335445
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,8,32,power_law_1.2,0.05017727851867676
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,2,128,power_law_1.2,1.1818316650390623
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,8,32,power_law_1.2,0.05363455772399902
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,8,32,power_law_1.2,0.05682432174682617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,8,32,power_law_1.2,0.056815361976623534
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,8,32,power_law_1.2,0.07179391860961915
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,8,32,power_law_1.2,0.08407936096191407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,8,32,power_law_1.2,0.09864831924438476
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,8,32,power_law_1.2,0.12344703674316407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,8,32,power_law_1.2,0.16418176651000976
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,8,32,power_law_1.2,0.2301568031311035
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,8,32,power_law_1.01,0.5630694580078125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,16,16,balanced,0.019738880395889283
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,16,16,balanced,0.01962623953819275
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,8,32,power_law_1.2,0.0339353609085083
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,16,16,balanced,0.021638400554656982
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,16,16,balanced,0.022984960079193116
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,16,16,balanced,0.03146368026733398
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,16,16,balanced,0.03231487989425659
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,8,32,power_law_1.2,0.37566078186035157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,16,16,balanced,0.032826879024505616
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,16,16,balanced,0.03312383890151978
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,16,16,balanced,0.03433216094970703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,16,16,balanced,0.03500159978866577
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,16,16,balanced,0.03607680082321167
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,16,16,balanced,0.036771841049194336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,16,16,balanced,0.03846528053283692
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,16,16,balanced,0.041475839614868164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,16,16,balanced,0.046163201332092285
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,16,16,balanced,0.04152703762054443
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,16,16,balanced,0.04899712085723877
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,16,16,balanced,0.05328383922576905
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,16,16,balanced,0.06920576095581055
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,16,16,balanced,0.07756415843963624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,16,16,balanced,0.09496447563171387
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,16,16,balanced,0.10808064460754394
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,16,16,balanced,0.142739200592041
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,16,16,balanced,0.17169279098510742
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,16,16,balanced,0.23408384323120118
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,16,16,power_law_1.01,0.021553919315338135
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,16,16,balanced,0.020341761112213134
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,16,16,balanced,0.2962137603759766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,16,16,power_law_1.01,0.0245580792427063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,16,16,power_law_1.01,0.024904959201812744
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,16,16,power_law_1.01,0.02496000051498413
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,16,16,power_law_1.01,0.030351359844207764
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,16,16,power_law_1.01,0.03147903919219971
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,16,16,power_law_1.01,0.03218303918838501
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,16,16,power_law_1.01,0.03468031883239746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,16,16,power_law_1.01,0.03689599990844726
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,8,32,power_law_1.2,0.4287820816040039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,16,16,power_law_1.01,0.037367041110992434
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,16,16,power_law_1.01,0.03829119920730591
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,16,16,power_law_1.01,0.041094398498535155
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,16,16,power_law_1.01,0.04836480140686035
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,16,16,power_law_1.01,0.04995711803436279
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,16,16,power_law_1.01,0.051324157714843756
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,16,16,power_law_1.01,0.0579315185546875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,16,16,power_law_1.01,0.06975743770599366
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,16,16,power_law_1.01,0.0741209602355957
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,16,16,power_law_1.01,0.08707327842712402
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,16,16,power_law_1.01,0.10264063835144044
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,16,16,power_law_1.01,0.13966976165771486
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,16,16,power_law_1.01,0.16835840225219728
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,16,16,power_law_1.01,0.23551488876342774
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,16,16,power_law_1.01,0.3122675132751465
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,16,16,power_law_1.01,0.43739265441894537
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,16,16,power_law_1.2,0.021529600620269776
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,16,16,power_law_1.01,0.5404172897338867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,16,16,power_law_1.2,0.023968000411987305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,16,16,power_law_1.2,0.024706559181213377
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,16,16,power_law_1.2,0.0318668794631958
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,16,16,power_law_1.2,0.030250239372253417
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,4,64,power_law_1.2,0.8457036590576171
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,16,16,power_law_1.01,0.03499648094177246
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,16,16,power_law_1.2,0.03268095970153809
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,16,16,power_law_1.2,0.0336678409576416
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,16,16,power_law_1.2,0.0364736008644104
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,16,16,power_law_1.2,0.03993599891662598
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,16,16,power_law_1.2,0.04143360137939453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,16,16,power_law_1.2,0.048703999519348146
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,16,16,power_law_1.2,0.052391681671142575
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,16,16,power_law_1.2,0.055900158882141116
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,16,16,power_law_1.2,0.06116608142852783
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,16,16,power_law_1.2,0.06888832092285156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,16,16,power_law_1.2,0.07986815929412841
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,16,16,power_law_1.2,0.024595201015472412
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,16,16,power_law_1.2,0.09469696044921874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,16,16,power_law_1.2,0.11460607528686524
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,16,16,power_law_1.2,0.1653696060180664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,16,16,power_law_1.2,0.19524480819702147
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,16,16,power_law_1.2,0.03479935884475708
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,16,16,power_law_1.2,0.03799295902252197
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,32,8,balanced,0.019276800155639647
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,32,8,balanced,0.020211200714111328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,32,8,balanced,0.020793600082397463
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,32,8,balanced,0.02448256015777588
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,8,32,power_law_1.2,0.6772236633300781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,32,8,balanced,0.027476480007171632
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,32,8,balanced,0.03610111951828003
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,32,8,balanced,0.03806335926055908
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,32,8,balanced,0.03881472110748291
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,32,8,balanced,0.039907839298248295
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,32,8,balanced,0.040881919860839847
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,32,8,balanced,0.04464767932891846
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,32,8,balanced,0.04816127777099609
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,32,8,balanced,0.04568064212799072
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,32,8,balanced,0.04863999843597412
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,32,8,balanced,0.05072512149810791
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,32,8,balanced,0.055147519111633306
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,32,8,balanced,0.06523263931274415
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,32,8,balanced,0.07355775833129882
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,32,8,balanced,0.09314047813415527
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,32,8,balanced,0.10522879600524901
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,16,16,power_law_1.2,0.30756095886230467
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,32,8,balanced,0.13610367774963378
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,32,8,balanced,0.160316162109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,32,8,balanced,0.2206643295288086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,32,8,balanced,0.2734003257751465
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,32,8,balanced,0.3849702453613281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,32,8,power_law_1.01,0.021127679347991944
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,32,8,balanced,0.4972787094116211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,32,8,power_law_1.01,0.024686079025268554
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,32,8,power_law_1.01,0.024808959960937498
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,32,8,balanced,0.037427198886871335
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,32,8,power_law_1.01,0.027453439235687254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,32,8,power_law_1.01,0.03470463991165161
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,32,8,power_law_1.01,0.03627264022827149
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,32,8,power_law_1.01,0.03719680070877075
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,32,8,power_law_1.01,0.03964416027069092
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,32,8,power_law_1.01,0.0415334415435791
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,32,8,power_law_1.01,0.04418176174163818
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,32,8,power_law_1.01,0.05432447910308837
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,16,16,power_law_1.2,0.5750617599487304
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,32,8,power_law_1.01,0.053738241195678715
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,32,8,power_law_1.01,0.054942722320556636
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,32,8,power_law_1.01,0.056270079612731935
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,32,8,power_law_1.01,0.06210944175720214
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,32,8,power_law_1.01,0.07186431884765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,32,8,power_law_1.01,0.026753280162811276
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,16,16,power_law_1.2,0.3948006439208984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,32,8,power_law_1.01,0.08336000442504883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,32,8,power_law_1.01,0.09722623825073243
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,32,8,power_law_1.01,0.11863679885864258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,32,8,power_law_1.01,0.03857664108276367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,32,8,power_law_1.01,0.1537446403503418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,32,8,power_law_1.01,0.16600448608398438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,32,8,power_law_1.01,0.24181247711181642
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,32,8,power_law_1.2,0.021038079261779787
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,32,8,power_law_1.2,0.023214080333709718
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,32,8,power_law_1.2,0.024524800777435303
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,32,8,power_law_1.01,0.4832294464111328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,32,8,power_law_1.2,0.02581376075744629
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,32,8,power_law_1.01,0.6064960098266601
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,32,8,power_law_1.2,0.027481598854064943
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,32,8,power_law_1.2,0.03472127914428711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,32,8,power_law_1.2,0.03641599893569946
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,32,8,power_law_1.2,0.03740799903869629
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,32,8,power_law_1.2,0.03881727933883667
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,32,8,power_law_1.2,0.039930880069732666
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,32,8,power_law_1.2,0.04255104064941406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,32,8,power_law_1.2,0.04527999877929688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,32,8,power_law_1.2,0.048211197853088375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,32,8,power_law_1.2,0.050206718444824217
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,32,8,power_law_1.2,0.05201151847839356
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,32,8,power_law_1.2,0.054814720153808595
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,32,8,power_law_1.2,0.060060157775878906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,32,8,power_law_1.01,0.3204889678955078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,32,8,power_law_1.2,0.07419007778167724
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,32,8,power_law_1.2,0.0837388801574707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,32,8,power_law_1.2,0.09960320472717285
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,32,8,power_law_1.2,0.12290816307067871
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,32,8,power_law_1.2,0.15546624183654784
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,32,8,power_law_1.2,0.20478847503662106
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,8,32,power_law_1.2,1.025753631591797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,1,256,balanced,0.05467391967773437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,1,256,balanced,0.054872322082519534
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,1,256,balanced,0.054638080596923835
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,1,256,balanced,0.06397823810577392
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,1,256,balanced,0.10583552360534668
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,16,16,power_law_1.2,0.8361280059814453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,32,8,power_law_1.2,0.27415679931640624
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,1,256,balanced,0.10655615806579591
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,1,256,balanced,0.10947327613830567
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,1,256,balanced,0.11184767723083495
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,1,256,balanced,0.11064959526062013
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,1,256,balanced,0.11312512397766113
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,1,256,balanced,0.11394816398620607
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,1,256,balanced,0.11375359535217285
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,1,256,balanced,0.1185331153869629
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,1,256,balanced,0.11978879928588868
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,1,256,balanced,0.13272319793701173
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,1,256,balanced,0.14087552070617676
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,1,256,balanced,0.14313344001770018
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,1,256,balanced,0.1466393566131592
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,1,256,balanced,0.16035200119018556
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,1,256,balanced,0.17502592086791993
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,1,256,balanced,0.191778564453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,32,8,power_law_1.2,0.38347007751464846
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,1,256,balanced,0.21128831863403322
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,1,256,balanced,0.29405567169189456
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,1,256,balanced,0.03349503993988037
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,1,256,balanced,0.3560678482055664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,1,256,balanced,0.03407232046127319
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,1,256,balanced,0.03374079942703247
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,1,256,balanced,0.5218278503417968
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,1,256,balanced,0.055487999916076655
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,1,256,balanced,0.055390720367431645
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,1,256,balanced,0.05649536132812501
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,1,256,balanced,0.05771647930145264
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,1,256,balanced,0.057608962059020996
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,1,256,balanced,0.057762560844421384
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,1,256,balanced,0.06103807926177979
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,1,256,balanced,0.0613478422164917
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,1,256,balanced,0.06252799987792969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,1,256,balanced,0.06765056133270264
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,1,256,balanced,0.06828159809112548
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,1,256,balanced,0.0736678409576416
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,1,256,balanced,0.10942208290100097
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,1,256,balanced,0.07913087844848633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,1,256,balanced,0.07706111907958985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,1,256,balanced,0.08503552436828613
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,1,256,balanced,0.042758398056030274
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,1,256,balanced,0.11584383964538573
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,1,256,balanced,0.12404992103576659
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,1,256,balanced,0.6341132736206054
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,1,256,balanced,0.14820863723754882
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,1,256,balanced,0.20050432205200197
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,1,256,balanced,0.24142080307006836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,32,8,power_law_1.2,0.5433190536499024
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,1,256,power_law_1.01,0.0639296007156372
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,1,256,balanced,0.34074241638183594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,1,256,balanced,0.4256614303588867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,1,256,power_law_1.01,0.06377984046936035
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,1,256,power_law_1.01,0.06371967792510987
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,1,256,power_law_1.01,0.10921216011047363
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,1,256,power_law_1.01,0.10534784317016602
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,1,256,power_law_1.01,0.11064191818237305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,1,256,power_law_1.01,0.112871675491333
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,1,256,power_law_1.01,0.12021504402160646
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,1,256,power_law_1.01,0.1114675235748291
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,1,256,balanced,0.09932671546936035
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,1,256,power_law_1.01,0.12340607643127441
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,1,256,power_law_1.01,0.13211392402648925
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,1,256,power_law_1.01,0.14225664138793945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,1,256,power_law_1.01,0.1474726390838623
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,1,256,power_law_1.01,0.15502464294433593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,1,256,power_law_1.01,0.16512767791748045
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,1,256,power_law_1.01,0.17724672317504883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,1,256,power_law_1.01,0.21170560836791993
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,1,256,power_law_1.01,0.2600793647766113
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,1,256,power_law_1.01,0.34570495605468754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,1,256,power_law_1.01,0.4716531372070312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,1,256,power_law_1.01,0.12527487754821778
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,1,256,power_law_1.01,0.6725286102294922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,1,256,power_law_1.01,0.9194866943359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,1,256,power_law_1.01,0.041715202331542967
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,1,256,power_law_1.01,0.04272895812988281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,1,256,power_law_1.01,0.04251008033752442
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,32,8,power_law_1.2,0.7422835540771484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,1,256,power_law_1.01,1.4929817199707032
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,1,256,power_law_1.01,0.05562111854553222
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,1,256,power_law_1.01,0.05674752235412598
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,1,256,power_law_1.01,0.05729407787322997
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,1,256,power_law_1.01,1.7278758239746093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,1,256,power_law_1.01,0.059953918457031255
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,1,256,power_law_1.01,0.06070911884307861
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,1,256,power_law_1.01,0.06511744022369384
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,1,256,power_law_1.01,0.06748032093048095
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,1,256,power_law_1.01,0.07058688163757323
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,1,256,power_law_1.01,0.07237247943878174
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,1,256,power_law_1.01,0.07567872047424316
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,1,256,power_law_1.01,0.08669568061828613
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,1,256,power_law_1.01,0.09264512062072754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,1,256,power_law_1.01,0.10488191604614258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,1,256,power_law_1.01,0.13267711639404295
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,1,256,power_law_1.01,0.12397439956665039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,1,256,power_law_1.01,0.17952640533447267
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,1,256,power_law_1.01,0.24192768096923828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,1,256,power_law_1.01,0.3040665626525879
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,1,256,power_law_1.01,2.8329690551757816
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,1,256,power_law_1.01,0.5264704132080078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,1,256,power_law_1.2,0.07046783924102783
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,1,256,power_law_1.01,0.8306956481933593
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,1,256,power_law_1.2,0.0639411211013794
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,1,256,power_law_1.2,0.06404863834381104
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,1,256,power_law_1.2,0.1061350440979004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,1,256,power_law_1.2,0.1100761604309082
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,1,256,power_law_1.2,0.10968704223632812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,1,256,power_law_1.01,1.3849562072753907
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,1,256,power_law_1.01,0.4091430282592773
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,1,256,power_law_1.2,0.11219327926635743
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,1,256,power_law_1.2,0.11638015747070311
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,1,256,power_law_1.01,4.122988891601563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,1,256,power_law_1.2,0.1302182388305664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,1,256,power_law_1.2,0.12375679969787598
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,1,256,power_law_1.2,0.1392793560028076
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,1,256,power_law_1.01,1.7459776306152341
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,1,256,power_law_1.2,0.1488499164581299
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,1,256,power_law_1.2,0.16508800506591798
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,1,256,power_law_1.2,0.17837568283081057
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,1,256,power_law_1.2,0.20280704498291016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,1,256,power_law_1.2,0.25100543975830075
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,1,256,power_law_1.2,0.3057855987548828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,1,256,power_law_1.01,2.3202406311035153
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,1,256,power_law_1.2,0.5096640014648438
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,1,256,power_law_1.2,0.12220159530639649
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,1,256,power_law_1.2,0.17454591751098633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,1,256,power_law_1.2,1.0342156982421875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,1,256,power_law_1.2,0.04239359855651855
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,1,256,power_law_1.2,1.2891891479492188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,1,256,power_law_1.2,0.04267263889312744
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,1,256,power_law_1.2,0.6164748764038086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,1,256,power_law_1.2,0.055302400588989255
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,1,256,power_law_1.2,0.05717504024505615
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,1,256,power_law_1.2,0.05761536121368408
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,1,256,power_law_1.2,0.06071424007415771
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,1,256,power_law_1.2,2.211270446777344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,1,256,power_law_1.2,0.06472320079803466
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,1,256,power_law_1.2,0.06613247871398927
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,1,256,power_law_1.2,0.042851839065551754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,1,256,power_law_1.2,0.07417600154876709
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,1,256,power_law_1.2,0.0736575984954834
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,1,256,power_law_1.2,0.07926144123077392
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,1,256,power_law_1.2,0.08927103996276856
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,1,256,power_law_1.2,0.10557184219360352
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,1,256,power_law_1.2,0.11868160247802735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,1,256,power_law_1.2,0.12861568450927735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,1,256,power_law_1.2,0.15614848136901854
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,1,256,power_law_1.2,3.11359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,1,256,power_law_1.2,0.18720640182495116
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,1,256,power_law_1.2,0.2762585639953613
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,1,256,power_law_1.2,0.3687027359008789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,1,256,power_law_1.2,0.06832384109497071
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,1,256,power_law_1.2,0.6071270370483399
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,2,128,balanced,0.025511679649353025
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,1,256,power_law_1.2,0.8085836791992188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,2,128,balanced,0.025470719337463376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,2,128,balanced,0.02634880065917969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,2,128,balanced,0.052851200103759766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,2,128,balanced,0.0926579189300537
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,1,256,power_law_1.2,1.2897894287109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,2,128,balanced,0.0932211208343506
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,2,128,balanced,0.09669376373291015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,2,128,balanced,0.0966374397277832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,1,256,power_law_1.2,1.8632281494140623
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,2,128,balanced,0.09843456268310546
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,1,256,power_law_1.2,5.185123901367187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,2,128,balanced,0.11008000373840332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,2,128,balanced,0.10387583732604981
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,2,128,balanced,0.09249279975891114
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,2,128,balanced,0.12321408271789551
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,2,128,balanced,0.11078271865844727
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,2,128,balanced,0.12152576446533203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,2,128,balanced,0.14157183647155763
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,2,128,balanced,0.15004799842834474
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,2,128,balanced,0.1510361576080322
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,1,256,power_law_1.2,3.0056436157226565
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,2,128,balanced,0.16888704299926757
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,2,128,balanced,0.21578880310058596
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,2,128,balanced,0.26210559844970704
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,2,128,balanced,0.3605324935913086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,2,128,balanced,0.10885503768920898
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,2,128,balanced,0.4344345474243164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,2,128,balanced,0.12536191940307617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,2,128,balanced,0.027118079662323
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,2,128,balanced,0.02721152067184448
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,2,128,balanced,0.03610879898071289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,2,128,balanced,0.05086336135864258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,2,128,balanced,0.6416320037841797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,2,128,balanced,0.05198463916778564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,2,128,balanced,0.05459968090057373
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,2,128,balanced,0.8073446655273437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,2,128,balanced,0.05458687782287598
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,2,128,balanced,0.057792000770568855
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,2,128,balanced,0.05871615886688233
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,2,128,balanced,0.06759935855865479
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,2,128,balanced,0.061399040222167967
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,2,128,balanced,0.06234752178192139
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,2,128,balanced,0.06830080032348633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,2,128,balanced,0.075065598487854
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,2,128,balanced,0.08772095680236816
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,2,128,balanced,0.10059391975402832
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,2,128,balanced,0.10265343666076661
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,2,128,balanced,0.12085503578186034
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,2,128,balanced,0.02715519905090332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,2,128,balanced,0.16002559661865234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,2,128,balanced,0.19362688064575195
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,2,128,balanced,0.2627264022827148
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,2,128,balanced,0.32385280609130856
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,2,128,power_law_1.01,0.05746431827545166
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,2,128,balanced,0.06359551906585693
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,2,128,balanced,0.47499649047851566
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,2,128,power_law_1.01,0.05707647800445557
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,2,128,balanced,0.611709442138672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,2,128,power_law_1.01,0.057839360237121586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,2,128,power_law_1.01,0.08910335540771484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,2,128,power_law_1.01,0.09077631950378418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,2,128,power_law_1.01,0.09423104286193848
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,2,128,power_law_1.01,0.0982374382019043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,2,128,power_law_1.01,0.10102144241333008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,2,128,power_law_1.01,0.11184767723083495
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,2,128,balanced,0.054504961967468256
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,2,128,power_law_1.01,0.11503104209899902
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,2,128,power_law_1.01,0.11580927848815918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,2,128,power_law_1.01,0.12622976303100586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,2,128,power_law_1.01,0.132741117477417
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,2,128,power_law_1.01,0.14789376258850098
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,2,128,power_law_1.01,0.16139904022216797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,1,256,power_law_1.2,7.499591674804688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,2,128,power_law_1.01,0.1890777587890625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,2,128,power_law_1.01,0.2233113670349121
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,2,128,power_law_1.01,0.3170240020751953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,2,128,power_law_1.01,0.382108154296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,2,128,power_law_1.01,0.09841535568237306
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,2,128,power_law_1.01,0.644579849243164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,2,128,power_law_1.01,0.9037171173095704
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,2,128,power_law_1.01,0.16629119873046877
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,2,128,power_law_1.01,0.040634880065917964
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,2,128,power_law_1.01,1.0789170837402344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,2,128,power_law_1.01,0.04241280078887939
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,2,128,power_law_1.01,0.04941567897796631
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,2,128,power_law_1.01,0.04948351860046386
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,2,128,power_law_1.01,0.05344895839691162
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,2,128,power_law_1.01,1.47304443359375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,2,128,power_law_1.01,0.0565388822555542
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,2,128,power_law_1.01,0.05990143775939941
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,2,128,power_law_1.01,0.0652620792388916
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,1,256,power_law_1.2,4.4212622070312495
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,2,128,power_law_1.01,0.06876800060272217
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,2,128,power_law_1.01,0.07505792140960693
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,2,128,power_law_1.01,0.0865113639831543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,2,128,power_law_1.01,0.09658623695373535
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,2,128,power_law_1.01,0.11125375747680663
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,2,128,power_law_1.01,2.474054412841797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,2,128,power_law_1.01,0.10720128059387206
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,2,128,power_law_1.01,0.15354751586914062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,2,128,power_law_1.01,0.06463744163513183
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,2,128,power_law_1.01,0.28469247817993165
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,2,128,power_law_1.01,0.4033638381958008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,2,128,power_law_1.01,3.236103820800781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,2,128,power_law_1.01,0.041218562126159666
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,2,128,power_law_1.01,0.5971366500854491
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,2,128,power_law_1.01,0.13223679542541505
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,2,128,power_law_1.01,0.7126322937011719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,2,128,power_law_1.2,0.057034239768981934
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,2,128,power_law_1.01,0.06399871826171874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,2,128,power_law_1.2,0.057067518234252934
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,2,128,power_law_1.2,0.057864961624145506
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,2,128,power_law_1.2,0.08836992263793945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,2,128,power_law_1.01,1.150492172241211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,2,128,power_law_1.2,0.0883238410949707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,2,128,power_law_1.2,0.09364480018615723
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,2,128,power_law_1.2,0.09645055770874024
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,2,128,power_law_1.2,0.10442111968994142
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,2,128,power_law_1.2,0.10669952392578126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,2,128,power_law_1.2,0.10504063606262207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,2,128,power_law_1.2,0.11347200393676757
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,2,128,power_law_1.2,0.11972352027893067
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,2,128,power_law_1.01,1.7303526306152341
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,2,128,power_law_1.2,0.1309452819824219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,2,128,power_law_1.01,0.22558208465576174
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,2,128,power_law_1.2,0.1431999969482422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,2,128,power_law_1.2,0.1531340789794922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,2,128,power_law_1.2,0.16381952285766602
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,2,128,power_law_1.2,0.21220352172851564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,2,128,power_law_1.2,0.16761728286743166
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,2,128,power_law_1.2,0.2504972839355469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,2,128,power_law_1.2,0.4233113479614258
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,2,128,power_law_1.01,2.2079321289062497
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,2,128,power_law_1.2,0.5660915374755859
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,2,128,power_law_1.2,0.7574079895019532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,2,128,power_law_1.2,0.04070784091949463
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,2,128,power_law_1.2,0.04123648166656494
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,2,128,power_law_1.2,0.04247424125671387
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,2,128,power_law_1.2,1.085011215209961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,2,128,power_law_1.2,0.04916096210479736
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,2,128,power_law_1.2,0.051280641555786134
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,2,128,power_law_1.2,1.4400575256347656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,2,128,power_law_1.2,0.053884158134460444
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,2,128,power_law_1.2,0.06185215950012207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,2,128,power_law_1.2,0.06466303825378418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,2,128,power_law_1.2,0.06585472106933594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,2,128,power_law_1.2,0.06335616111755371
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,2,128,power_law_1.2,0.07360000133514404
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,2,128,power_law_1.2,0.07574272155761719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,2,128,power_law_1.2,0.0832153606414795
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,2,128,power_law_1.2,0.0920358371734619
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,2,128,power_law_1.2,0.10888192176818848
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,2,128,power_law_1.2,0.11269887924194336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,2,128,power_law_1.2,0.15410176277160645
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,2,128,power_law_1.2,2.21204345703125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,2,128,power_law_1.2,0.17102720260620116
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,2,128,power_law_1.2,0.30385408401489256
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,2,128,power_law_1.2,0.3315302276611328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,2,128,power_law_1.2,0.054876160621643064
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,2,128,power_law_1.2,0.5414400100708008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,4,64,balanced,0.023573760986328123
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,2,128,power_law_1.2,0.6986099243164062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,4,64,balanced,0.03666431903839111
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,4,64,balanced,0.04293248176574707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,4,64,balanced,0.05398399829864502
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,2,128,power_law_1.2,1.094961929321289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,4,64,balanced,0.08770815849304199
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,2,128,power_law_1.2,3.5848165893554684
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,4,64,balanced,0.08998016357421876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,4,64,balanced,0.09850367546081543
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,4,64,balanced,0.09401344299316407
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,4,64,balanced,0.10919936180114746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,4,64,balanced,0.09837311744689942
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,4,64,balanced,0.10065279960632326
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,4,64,balanced,0.10360960006713867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,2,128,power_law_1.2,1.4540594482421876
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,4,64,balanced,0.1102412796020508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,4,64,balanced,0.11556096076965332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,4,64,balanced,0.11824511528015136
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,2,128,power_law_1.2,4.308760375976563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,4,64,balanced,0.1289241600036621
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,4,64,balanced,0.13613696098327638
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,4,64,balanced,0.13674367904663084
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,4,64,balanced,0.1505497646331787
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,4,64,balanced,0.18093696594238282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,4,64,balanced,0.22417280197143558
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,4,64,balanced,0.33840896606445314
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,4,64,balanced,0.4880409622192383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,4,64,balanced,0.025681920051574707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,4,64,balanced,0.615750389099121
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,4,64,balanced,0.026753280162811276
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,4,64,balanced,0.03757312059402466
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,4,64,balanced,0.05355519771575927
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,4,64,balanced,0.9062540435791016
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,2,128,power_law_1.2,2.8622311401367186
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,4,64,balanced,0.0562713623046875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,4,64,balanced,0.060684800148010254
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,4,64,balanced,0.06433536052703857
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,4,64,balanced,0.07057792186737061
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,4,64,balanced,0.06074495792388916
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,4,64,balanced,0.06552320003509522
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,4,64,balanced,0.2822233581542969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,4,64,balanced,1.1694989013671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,4,64,balanced,0.06320767879486085
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,2,128,power_law_1.2,2.099886016845703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,4,64,balanced,0.07232768058776855
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,4,64,balanced,0.07694719791412354
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,4,64,balanced,0.08363391876220702
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,4,64,balanced,0.09406592369079589
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,4,64,balanced,0.1104703998565674
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,4,64,balanced,0.13947263717651368
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,4,64,balanced,0.16928512573242188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,4,64,balanced,0.02773632049560547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,4,64,balanced,0.2294207954406738
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,4,64,balanced,0.28356992721557617
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,4,64,balanced,0.41918079376220707
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,4,64,power_law_1.01,0.05700223922729493
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,4,64,balanced,0.5310438537597657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,4,64,balanced,0.06762239933013917
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,4,64,balanced,0.7585139465332031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,4,64,power_law_1.01,0.05750271797180175
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,4,64,power_law_1.01,0.05889664173126221
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,4,64,balanced,0.09724160194396972
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,4,64,balanced,1.0054579162597657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,4,64,power_law_1.01,0.0844876766204834
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,4,64,power_law_1.01,0.09362815856933594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,4,64,power_law_1.01,0.08848896026611328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,4,64,power_law_1.01,0.10811776161193846
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,4,64,power_law_1.01,0.10920831680297852
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,4,64,power_law_1.01,0.11079296112060547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,4,64,power_law_1.01,0.1149721622467041
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,4,64,power_law_1.01,0.12813055992126465
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,4,64,power_law_1.01,0.13671936035156249
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,4,64,power_law_1.01,0.1480140781402588
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,4,64,power_law_1.01,0.147705602645874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,4,64,power_law_1.01,0.17792640686035158
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,4,64,power_law_1.01,0.2079795265197754
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,4,64,power_law_1.01,0.3070899200439453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,4,64,power_law_1.01,0.33557376861572263
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,4,64,power_law_1.01,0.09738752365112305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,4,64,power_law_1.01,0.5145151901245117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,4,64,power_law_1.01,0.10634495735168456
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,4,64,power_law_1.01,0.06020607948303223
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,4,64,power_law_1.01,0.7338355255126954
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,4,64,power_law_1.01,0.03991039991378784
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,4,64,power_law_1.01,0.04079360008239746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,4,64,power_law_1.01,0.042535681724548344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,4,64,power_law_1.01,0.052508158683776854
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,4,64,power_law_1.01,0.055997438430786126
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,4,64,power_law_1.01,1.3148825073242187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,4,64,power_law_1.01,0.06918911933898926
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,4,64,power_law_1.01,0.05771135807037353
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,4,64,power_law_1.01,0.06818560123443604
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,4,64,power_law_1.01,0.06319615840911866
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,4,64,power_law_1.01,0.06753024101257324
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,4,64,power_law_1.01,0.03905280113220215
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,4,64,power_law_1.01,0.07664127826690674
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,4,64,power_law_1.01,0.0834175968170166
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,4,64,power_law_1.01,0.09899904251098632
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,4,64,power_law_1.01,0.10300671577453613
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,4,64,power_law_1.01,1.8984640502929686
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,4,64,power_law_1.01,0.10733311653137208
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,4,64,power_law_1.01,0.134584321975708
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,4,64,power_law_1.01,1.1082905578613282
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,4,64,power_law_1.01,0.1601638412475586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,4,64,power_law_1.01,0.205533447265625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,4,64,power_law_1.01,0.26280832290649414
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,4,64,power_law_1.01,0.4527731323242188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,4,64,power_law_1.01,2.8119552612304686
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,4,64,power_law_1.01,0.5099532699584961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,4,64,power_law_1.01,0.07213952064514159
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,4,64,power_law_1.2,0.056624641418457036
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,4,64,power_law_1.2,0.06391935825347901
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,4,64,power_law_1.01,0.8870642852783203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,4,64,power_law_1.01,0.9947392272949219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,4,64,power_law_1.2,0.08756480216979981
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,4,64,power_law_1.2,0.09094911575317383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,4,64,power_law_1.2,0.09049471855163574
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,4,64,power_law_1.2,0.10286720275878906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,4,64,power_law_1.2,0.10724736213684083
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,4,64,power_law_1.2,0.1051750373840332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,4,64,power_law_1.01,1.4909555053710937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,4,64,power_law_1.2,0.10874367713928224
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,4,64,power_law_1.2,0.11475584030151367
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,4,64,power_law_1.2,0.12485247611999513
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,4,64,power_law_1.2,0.1360102367401123
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,4,64,power_law_1.2,0.13987327575683595
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,4,64,power_law_1.2,0.058154239654541015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,4,64,power_law_1.2,0.15255680084228515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,4,64,power_law_1.2,0.19212928771972657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,4,64,power_law_1.01,1.8630680847167969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,4,64,power_law_1.2,0.2359424018859863
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,4,64,power_law_1.2,0.3261503982543945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,4,64,power_law_1.2,0.46615936279296877
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,4,64,power_law_1.2,0.6533849334716797
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,4,64,power_law_1.2,0.8654771423339843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,4,64,power_law_1.2,0.039744000434875484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,4,64,power_law_1.2,0.060209918022155764
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,4,64,power_law_1.2,0.039888639450073246
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,4,64,power_law_1.2,0.04129280090332031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,4,64,power_law_1.2,1.274956817626953
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,4,64,power_law_1.2,0.042617602348327635
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,4,64,power_law_1.2,0.05175424098968506
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,4,64,power_law_1.2,0.06536960124969482
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,4,64,power_law_1.2,1.720506896972656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,4,64,power_law_1.2,0.06135551929473877
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,4,64,power_law_1.2,0.062492160797119146
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,4,64,power_law_1.2,0.06798208236694336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,4,64,power_law_1.2,0.14401920318603517
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,4,64,power_law_1.2,0.07482495784759521
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,4,64,power_law_1.2,0.0772492790222168
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,4,64,power_law_1.2,0.09357952117919922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,4,64,power_law_1.2,0.10131967544555665
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,4,64,power_law_1.2,0.10048512458801269
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,4,64,power_law_1.2,0.11644800186157227
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,4,64,power_law_1.2,0.05703680038452148
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,4,64,power_law_1.2,0.14249216079711916
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,4,64,power_law_1.2,0.17595647811889648
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,4,64,power_law_1.2,2.7539071655273437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,4,64,power_law_1.2,0.06407423973083495
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,4,64,power_law_1.2,0.2813759994506836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,4,64,power_law_1.2,0.34838783264160156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,4,64,power_law_1.2,0.49150848388671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,8,32,balanced,0.02313215970993042
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,4,64,power_law_1.2,0.6485145568847657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,8,32,balanced,0.03145600080490112
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,8,32,balanced,0.04197120189666748
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,8,32,balanced,0.05406079769134522
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,4,64,power_law_1.2,0.945132827758789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,8,32,balanced,0.09521535873413087
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,8,32,balanced,0.11042431831359863
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,8,32,balanced,0.10053759574890138
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,8,32,balanced,0.10124287605285645
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,4,64,power_law_1.2,1.3660671997070313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,8,32,balanced,0.11048064231872559
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,8,32,balanced,0.11296640396118166
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,8,32,balanced,0.12377856254577638
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,8,32,balanced,0.12697728157043459
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,8,32,balanced,0.13459712028503418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,8,32,balanced,0.14661375999450682
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,8,32,balanced,0.14320384025573732
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,8,32,balanced,0.16993535995483397
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,8,32,balanced,0.06639359951019287
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,8,32,balanced,0.1948966407775879
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,8,32,balanced,0.2508620834350586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,8,32,balanced,0.2992831993103027
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,8,32,balanced,0.4198118209838867
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,8,32,balanced,0.5242547225952149
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,4,64,power_law_1.2,3.7494976806640623
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,8,32,balanced,0.13273856163024902
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,8,32,balanced,0.7740940856933594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,8,32,balanced,0.02610431909561157
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,8,32,balanced,0.02515455961227417
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,8,32,balanced,0.028275198936462402
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,8,32,balanced,0.9849971008300781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,8,32,balanced,0.03815551996231079
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,8,32,balanced,0.04950528144836426
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,8,32,balanced,0.06523263931274415
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,4,64,power_law_1.2,1.9554649353027345
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,8,32,balanced,0.07363711833953858
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,8,32,balanced,0.06361087799072265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,8,32,balanced,0.06443647861480713
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,8,32,balanced,0.06952191829681396
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,8,32,balanced,0.0748044776916504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,8,32,balanced,0.08301440238952637
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,8,32,balanced,0.09546367645263672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,8,32,balanced,0.10782208442687988
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,8,32,balanced,1.4588262939453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,8,32,balanced,0.10548352241516112
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,8,32,balanced,0.1315494441986084
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,8,32,balanced,0.1584716796875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,8,32,balanced,0.20707712173461915
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,8,32,balanced,0.2555609512329101
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,8,32,balanced,0.3622643280029297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,8,32,balanced,1.911546936035156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,8,32,balanced,0.4612812805175781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,8,32,balanced,0.0766707181930542
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,8,32,power_law_1.01,0.05928703784942627
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,8,32,balanced,0.6843814086914063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,8,32,balanced,0.09400064468383788
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,8,32,balanced,0.8791104125976563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,8,32,power_law_1.01,0.05903872013092041
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,8,32,power_law_1.01,0.059123201370239256
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,8,32,power_law_1.01,0.06232831954956055
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,8,32,power_law_1.01,0.06672383785247803
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,8,32,power_law_1.01,0.10022912025451661
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,8,32,power_law_1.01,0.09015040397644043
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,8,32,power_law_1.01,0.10783359527587891
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,8,32,balanced,1.2844671630859374
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,8,32,power_law_1.01,0.11039360046386719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,8,32,power_law_1.01,0.1155776023864746
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,8,32,power_law_1.01,0.1283520030975342
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,8,32,power_law_1.01,0.13514880180358885
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,8,32,power_law_1.01,0.14138496398925782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,8,32,power_law_1.01,0.13988608360290528
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,8,32,balanced,1.7045542907714843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,8,32,power_law_1.01,0.1540403175354004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,8,32,power_law_1.01,0.1914227294921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,8,32,power_law_1.01,0.21778944015502927
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,4,64,power_law_1.2,2.9412017822265626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,8,32,power_law_1.01,0.2959974479675293
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,8,32,power_law_1.01,0.3416844940185547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,8,32,power_law_1.01,0.10915200233459472
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,8,32,power_law_1.01,0.5372851181030274
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,8,32,power_law_1.01,0.04230144023895264
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,8,32,power_law_1.01,0.7225740814208985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,8,32,power_law_1.01,0.04294400215148926
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,8,32,power_law_1.01,0.04369919776916504
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,8,32,power_law_1.01,1.0977740478515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,8,32,power_law_1.01,0.045114879608154294
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,8,32,power_law_1.01,0.10042880058288575
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,8,32,power_law_1.01,1.347527618408203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,8,32,power_law_1.01,0.048981761932373045
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,8,32,power_law_1.01,0.06814080238342285
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,8,32,power_law_1.01,0.07059328079223633
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,8,32,power_law_1.01,0.06456319808959961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,8,32,power_law_1.01,0.0732531213760376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,8,32,power_law_1.01,0.07856256008148194
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,8,32,power_law_1.01,0.08231679916381836
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,8,32,power_law_1.01,0.09539072036743164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,8,32,power_law_1.01,2.116636199951172
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,8,32,power_law_1.01,0.10226304054260253
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,8,32,power_law_1.01,0.09742207527160644
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,8,32,power_law_1.01,0.11384448051452636
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,8,32,power_law_1.01,0.18034175872802732
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,8,32,power_law_1.01,0.06236671924591064
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,8,32,power_law_1.01,0.23320703506469725
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,8,32,power_law_1.01,0.2764761543273926
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,8,32,power_law_1.01,0.06542719841003418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,8,32,power_law_1.01,2.846188659667969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,8,32,power_law_1.01,0.48474494934082035
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,8,32,power_law_1.01,0.5869452667236328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,8,32,power_law_1.2,0.059024639129638665
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,8,32,power_law_1.01,0.8006259155273436
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,8,32,power_law_1.01,0.15193471908569336
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,8,32,power_law_1.2,0.058952960968017575
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,8,32,power_law_1.2,0.059321599006652834
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,8,32,power_law_1.01,1.0929408264160156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,8,32,power_law_1.2,0.06569983959197997
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,8,32,power_law_1.2,0.1064089584350586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,8,32,power_law_1.2,0.09209600448608399
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,8,32,power_law_1.2,0.10761856079101562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,8,32,power_law_1.2,0.10411647796630859
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,8,32,power_law_1.2,0.11733247756958007
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,8,32,power_law_1.01,1.5850086975097657
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,8,32,power_law_1.2,0.13110272407531737
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,8,32,power_law_1.2,0.14676863670349122
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,8,32,power_law_1.2,0.14549375534057618
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,8,32,power_law_1.2,0.1661747169494629
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,8,32,power_law_1.2,0.20196479797363281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,8,32,power_law_1.2,0.15439104080200194
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,8,32,power_law_1.2,0.23240703582763672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,8,32,power_law_1.2,0.06300928115844726
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,8,32,power_law_1.2,0.10159616470336914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,8,32,power_law_1.01,2.3279577636718747
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,8,32,power_law_1.2,0.3376063919067383
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,8,32,power_law_1.2,0.4393331146240234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,8,32,power_law_1.2,0.6475750732421874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,8,32,power_law_1.2,0.04195199966430664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,8,32,power_law_1.2,0.8426278686523437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,8,32,power_law_1.2,0.042986240386962894
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,8,32,power_law_1.2,0.043541760444641114
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,8,32,power_law_1.2,1.191514892578125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,8,32,power_law_1.2,0.04512256145477295
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,8,32,power_law_1.2,0.04799871921539307
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,8,32,power_law_1.2,0.06144000053405761
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,8,32,power_law_1.2,0.12213631629943848
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,8,32,power_law_1.2,0.07164031982421874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,8,32,power_law_1.2,0.06605440139770508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,8,32,power_law_1.2,1.5607589721679687
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,8,32,power_law_1.2,0.06723328113555908
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,8,32,power_law_1.2,0.07862400054931641
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,8,32,power_law_1.2,0.08339839935302734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,8,32,power_law_1.2,0.10528767585754395
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,8,32,power_law_1.2,0.101396484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,8,32,power_law_1.2,0.1076352024078369
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,8,32,power_law_1.2,0.12405247688293457
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,8,32,power_law_1.2,0.1493900775909424
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,8,32,power_law_1.2,0.1830899238586426
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,8,32,power_law_1.2,2.4286656188964844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,8,32,power_law_1.2,0.2595417594909668
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,8,32,power_law_1.2,0.30771200180053715
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,8,32,power_law_1.2,0.06938240051269531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,8,32,power_law_1.2,0.07700863838195801
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,8,32,power_law_1.2,0.49187969207763677
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,8,32,power_law_1.2,0.5944524765014648
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,16,16,balanced,0.04769536018371582
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,16,16,balanced,0.06624127864837646
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,8,32,power_law_1.2,0.9468863677978515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,16,16,balanced,0.09068799972534179
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,16,16,balanced,0.10886655807495119
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,8,32,power_law_1.2,1.2683609771728517
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,16,16,balanced,0.13291135787963868
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,16,16,balanced,0.13568127632141114
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,16,16,balanced,0.14709504127502443
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,16,16,balanced,0.14670080184936524
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,16,16,balanced,0.1535961627960205
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,16,16,balanced,0.02603775978088379
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,16,16,balanced,0.16382591247558592
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,16,16,balanced,0.16876544952392578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,16,16,balanced,0.18629632949829103
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,16,16,balanced,0.16970752716064452
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,8,32,power_law_1.2,1.9716531372070314
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,16,16,balanced,0.18031744003295896
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,16,16,balanced,0.2005465507507324
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,16,16,balanced,0.24181119918823243
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,16,16,balanced,0.28442623138427736
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,16,16,balanced,0.4055449676513672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,16,16,balanced,0.4848640060424804
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,16,16,balanced,0.03171583890914917
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,16,16,balanced,0.6849446105957031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,8,32,power_law_1.2,3.146134948730469
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,16,16,balanced,0.8463385772705078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,16,16,balanced,0.10036735534667969
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,16,16,balanced,0.13421952247619628
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,16,16,balanced,0.2172966384887695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,16,16,balanced,1.264337921142578
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,16,16,balanced,0.3864064025878906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,16,16,balanced,1.6355148315429688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,16,16,balanced,0.649591064453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,16,16,balanced,0.5784767913818359
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,16,16,balanced,0.746240005493164
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,16,16,balanced,0.5850048065185547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,16,16,balanced,0.5775884628295899
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,16,16,balanced,2.408030700683594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,16,16,balanced,0.5889625549316406
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,16,16,balanced,0.5949004745483398
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,16,16,balanced,0.6012377548217773
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,16,16,balanced,0.6128729629516602
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,16,16,balanced,0.6313625717163086
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,16,16,balanced,0.6292915344238281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,16,16,balanced,0.6132608032226562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,16,16,balanced,0.6516083526611328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,16,16,balanced,3.129564208984375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,16,16,balanced,0.6947634887695313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,16,16,balanced,0.7477977752685547
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,16,16,balanced,0.8358246612548829
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,8,32,power_law_1.2,2.650025024414062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,16,16,balanced,0.9334976196289062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,16,16,power_law_1.01,0.0768012809753418
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,16,16,power_law_1.01,0.08879743576049805
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,16,16,power_law_1.01,0.08710271835327148
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,16,16,balanced,1.6108697509765626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,16,16,power_law_1.01,0.09936639785766602
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,16,16,power_law_1.01,0.09377920150756835
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,16,16,balanced,1.8271897888183595
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,16,16,power_law_1.01,0.10299903869628906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,16,16,power_law_1.01,0.13367679595947266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,16,16,power_law_1.01,0.13387904167175294
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,16,16,power_law_1.01,0.13917311668395996
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,16,16,power_law_1.01,0.1458777618408203
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,16,16,power_law_1.01,0.1518617630004883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,16,16,power_law_1.01,0.1594752025604248
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,16,16,balanced,2.698506164550781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,16,16,power_law_1.01,0.17841535568237304
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,16,16,power_law_1.01,0.1632102394104004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,16,16,power_law_1.01,0.17439104080200196
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,16,16,power_law_1.01,0.24316160202026368
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,16,16,power_law_1.01,0.29957759857177735
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,16,16,balanced,3.0714804077148434
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,16,16,power_law_1.01,0.13861120223999024
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,16,16,power_law_1.01,0.38766590118408206
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,16,16,power_law_1.01,0.48855041503906255
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,16,16,power_law_1.01,0.7020928192138671
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,16,16,power_law_1.01,0.9233369445800781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,16,16,power_law_1.01,1.3261798095703123
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,16,16,balanced,4.815112915039062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,16,16,power_law_1.01,0.5169740676879883
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,16,16,power_law_1.01,1.7501785278320312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,16,16,power_law_1.01,0.19643135070800782
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,16,16,power_law_1.01,0.7024217224121093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,16,16,power_law_1.01,0.7201651000976563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,16,16,balanced,6.131484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,16,16,power_law_1.01,2.4887744140625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,16,16,power_law_1.01,0.7247987365722656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,16,16,power_law_1.01,0.57478271484375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,16,16,power_law_1.01,0.5546355056762695
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,16,16,power_law_1.01,0.5652108764648437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,16,16,power_law_1.01,0.5588390350341796
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,16,16,power_law_1.01,0.5710796737670899
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,16,16,power_law_1.01,0.5757759857177734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,16,16,power_law_1.01,3.281675415039062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,16,16,power_law_1.01,0.5813350296020507
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,16,16,power_law_1.01,0.5905484771728515
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,16,16,power_law_1.01,0.6090521621704101
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,16,16,power_law_1.01,0.6069631958007813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,16,16,power_law_1.01,0.6343577575683594
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,16,16,power_law_1.01,0.7668057250976562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,16,16,power_law_1.01,0.8615347290039063
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,16,16,power_law_1.01,1.0955648040771484
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,16,16,power_law_1.01,1.314001922607422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,16,16,power_law_1.01,1.8164274597167966
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,16,16,power_law_1.2,0.07177728176116943
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,16,16,power_law_1.01,2.276873016357422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,16,16,power_law_1.2,0.09023743629455566
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,16,16,power_law_1.2,0.09308927536010743
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,16,16,power_law_1.01,3.2145214843749996
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,16,16,power_law_1.2,0.09914239883422851
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,16,16,power_law_1.2,0.13738368034362794
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,16,16,power_law_1.01,3.975166625976563
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,16,16,power_law_1.2,0.1346944046020508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,16,16,power_law_1.2,0.08463744163513184
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,16,16,power_law_1.2,0.1386854362487793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,16,16,power_law_1.2,0.14583935737609863
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,16,16,power_law_1.2,0.15273983955383302
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,16,16,power_law_1.2,0.16372735977172853
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,16,16,power_law_1.2,0.12949376106262206
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,16,16,power_law_1.2,0.18393472671508787
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,16,16,power_law_1.2,0.16928768157958984
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,16,16,power_law_1.2,0.10784895896911621
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,16,16,power_law_1.2,0.18202367782592774
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,16,16,power_law_1.2,0.20276607513427733
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,16,16,power_law_1.01,5.6116162109375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,16,16,power_law_1.2,0.24882816314697265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,16,16,power_law_1.2,0.3037887954711914
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,16,16,power_law_1.2,0.41916545867919924
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,16,16,power_law_1.2,0.5306905746459961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,16,16,power_law_1.2,0.7596390533447266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,16,16,power_law_1.2,1.0061427307128905
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,16,16,power_law_1.01,7.491171875000001
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,16,16,power_law_1.2,1.4749964904785156
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,16,16,power_law_1.2,0.519444465637207
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,16,16,power_law_1.2,1.8772210693359377
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,16,16,power_law_1.2,0.653802261352539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,16,16,power_law_1.2,0.6880127716064453
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,16,16,power_law_1.2,0.707325439453125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,16,16,power_law_1.2,0.568935661315918
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,16,16,power_law_1.2,0.7239910125732422
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,16,16,power_law_1.2,0.7408614349365235
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,16,16,power_law_1.2,2.813553771972656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,16,16,power_law_1.2,0.5617881774902344
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,16,16,power_law_1.01,0.7007782745361328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,16,16,power_law_1.2,0.5607027053833008
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,16,16,power_law_1.2,0.5660236740112305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,16,16,power_law_1.2,0.5679244613647461
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,16,16,power_law_1.01,0.7074317169189454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,16,16,power_law_1.2,0.5832255935668945
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,16,16,power_law_1.2,0.5949491119384765
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,16,16,power_law_1.2,0.6156377410888672
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,16,16,power_law_1.2,0.5943756866455078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,16,16,power_law_1.2,0.64614013671875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,16,16,power_law_1.2,0.7119475555419922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,16,16,power_law_1.2,0.7904000091552734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,16,16,power_law_1.2,0.915389404296875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,16,16,power_law_1.2,1.1747494506835938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,16,16,power_law_1.2,1.4543463134765626
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,32,8,balanced,0.026936318874359134
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,32,8,balanced,0.03319295883178711
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,16,16,power_law_1.2,1.898662414550781
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,32,8,balanced,0.07134335994720459
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,32,8,balanced,0.09938816070556641
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,32,8,balanced,0.12456576347351074
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,16,16,power_law_1.2,2.565551452636719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,32,8,balanced,0.13638015747070314
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,32,8,balanced,0.1418291187286377
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,16,16,power_law_1.2,3.9566976928710935
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,32,8,balanced,0.1541337585449219
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,32,8,balanced,0.15684224128723145
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,32,8,balanced,0.17215488433837892
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,32,8,balanced,0.1787980842590332
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,16,16,power_law_1.2,3.517578125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,32,8,balanced,0.19228544235229492
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,32,8,balanced,0.05397759914398194
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,32,8,balanced,0.18728063583374024
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,32,8,balanced,0.20755584716796874
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,32,8,balanced,0.24419200897216795
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,32,8,balanced,0.2818137550354004
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,32,8,balanced,0.4126476669311524
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,32,8,balanced,0.16372608184814452
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,32,8,balanced,0.4876556777954102
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,32,8,balanced,0.17941120147705078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,16,16,power_law_1.2,4.853568115234375
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,32,8,balanced,0.7345779418945313
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,32,8,balanced,0.08256256103515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,32,8,balanced,0.8862937927246094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,32,8,balanced,0.13495936393737792
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,32,8,balanced,1.3002406311035155
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,32,8,balanced,0.40625537872314454
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,32,8,balanced,1.7033407592773435
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,32,8,balanced,0.637918701171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,32,8,balanced,0.8579853057861329
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,32,8,balanced,0.8277478027343751
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,16,16,power_law_1.2,6.887420043945312
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,32,8,balanced,2.511871948242187
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,32,8,balanced,0.8225778961181641
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,32,8,balanced,0.8024908447265625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,32,8,balanced,0.79963134765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,32,8,balanced,3.226278381347656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,32,8,balanced,0.8149644470214843
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,32,8,balanced,0.8203852844238281
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,32,8,balanced,0.8442803192138673
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,32,8,balanced,0.8263385772705079
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,32,8,balanced,0.8364672088623047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,32,8,balanced,0.88155517578125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,32,8,balanced,0.9281983947753906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,32,8,balanced,0.8190412902832032
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,32,8,balanced,0.9741363525390625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,32,8,balanced,1.0782911682128906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,32,8,balanced,1.1673715209960938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,32,8,balanced,0.2201241683959961
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,32,8,power_law_1.01,0.059797759056091304
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,32,8,balanced,2.0586419677734376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,32,8,balanced,2.25728515625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,32,8,power_law_1.01,0.0886899185180664
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,32,8,power_law_1.01,0.09263872146606446
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,32,8,power_law_1.01,0.10680959701538087
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,32,8,balanced,3.3287628173828123
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,32,8,balanced,3.8040319824218747
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,32,8,power_law_1.01,0.08364543914794922
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,32,8,power_law_1.01,0.13461376190185548
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,16,16,power_law_1.2,9.598072509765625
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,32,8,power_law_1.01,0.121146879196167
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,32,8,power_law_1.01,0.1302451229095459
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,32,8,balanced,5.90758544921875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,32,8,power_law_1.01,0.14940287590026854
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,32,8,power_law_1.01,0.1877235221862793
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,32,8,power_law_1.01,0.15224960327148437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,32,8,power_law_1.01,0.16556928634643556
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,32,8,power_law_1.01,0.21016704559326174
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,32,8,power_law_1.01,0.15433088302612305
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,32,8,power_law_1.01,0.26043775558471677
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,32,8,power_law_1.01,0.3087027168273926
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,32,8,balanced,7.455160522460938
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,32,8,power_law_1.01,0.4099942398071289
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,32,8,power_law_1.01,0.503298568725586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,32,8,power_law_1.01,0.16892288208007813
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,32,8,power_law_1.01,0.7522726440429688
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,32,8,power_law_1.01,0.186744327545166
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,32,8,power_law_1.01,0.9234815979003906
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,32,8,power_law_1.01,0.3241177749633789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,32,8,power_law_1.01,0.17242368698120117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,32,8,power_law_1.01,0.5999692916870117
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,32,8,power_law_1.01,0.4771699142456055
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,32,8,power_law_1.01,0.6019404983520508
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,32,8,power_law_1.01,1.7769612121582032
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,32,8,power_law_1.01,0.7355609893798828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,32,8,power_law_1.01,0.7650406646728516
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,32,8,power_law_1.01,0.7707302093505859
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,32,8,power_law_1.01,2.6583065795898437
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,32,8,power_law_1.01,3.575811767578125
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,32,8,power_law_1.01,1.3523762512207032
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,32,8,power_law_1.01,0.8587776184082031
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,32,8,power_law_1.01,0.7989401245117188
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,32,8,power_law_1.01,0.8311283111572265
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,32,8,power_law_1.01,0.980047378540039
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,32,8,power_law_1.01,1.088015365600586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,32,8,power_law_1.01,1.359936065673828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,32,8,power_law_1.01,0.7764710235595703
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,32,8,power_law_1.01,0.7385651397705078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,32,8,power_law_1.01,0.7535001373291015
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,32,8,power_law_1.01,1.5828402709960936
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,32,8,power_law_1.01,0.7729702758789062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,32,8,power_law_1.01,0.7799193572998047
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,32,8,power_law_1.01,0.806277084350586
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,32,8,power_law_1.01,2.121227569580078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,32,8,power_law_1.01,0.7686233520507812
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,32,8,power_law_1.2,0.06003071784973144
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,32,8,power_law_1.2,0.07350143909454346
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,32,8,power_law_1.2,0.09669247627258301
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,32,8,power_law_1.01,2.6408731079101564
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,32,8,power_law_1.2,0.10843775749206544
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,32,8,power_law_1.2,0.11541888236999512
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,32,8,power_law_1.2,0.1277132797241211
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,32,8,power_law_1.2,0.13620991706848146
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,32,8,power_law_1.2,0.14694527626037596
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,32,8,power_law_1.2,0.1504089641571045
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,32,8,power_law_1.2,0.15567999839782715
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,32,8,power_law_1.2,0.16718591690063475
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,32,8,power_law_1.2,0.17307136535644532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,32,8,power_law_1.2,0.19007871627807618
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,32,8,power_law_1.2,0.1762380790710449
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,32,8,power_law_1.01,3.5745278930664064
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,32,8,power_law_1.2,0.18740991592407225
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,32,8,power_law_1.2,0.08429056167602539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,32,8,power_law_1.2,0.21543039321899413
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,32,8,power_law_1.2,0.26716543197631837
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,32,8,power_law_1.2,0.3155878448486328
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,32,8,power_law_1.2,0.4130316925048828
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,32,8,power_law_1.2,0.5108902359008789
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,32,8,power_law_1.2,0.7606246185302734
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,32,8,power_law_1.01,4.678721313476562
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,32,8,power_law_1.2,0.9392396545410155
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,32,8,power_law_1.2,0.3259955215454101
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,32,8,power_law_1.2,0.42685184478759763
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,32,8,power_law_1.2,1.471998748779297
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,32,8,power_law_1.2,1.8437426757812498
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,32,8,power_law_1.2,0.5298406219482421
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,32,8,power_law_1.2,0.6290188980102539
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,32,8,power_law_1.2,0.7325004577636719
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,32,8,power_law_1.01,6.818447265624999
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,32,8,power_law_1.2,0.7500863647460937
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,32,8,power_law_1.2,0.7571097564697266
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,32,8,power_law_1.2,2.78951171875
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,32,8,power_law_1.2,0.7448639678955078
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,32,8,power_law_1.2,0.7403008270263671
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,32,8,power_law_1.2,0.7533299255371093
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,32,8,power_law_1.2,0.7670079803466796
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,32,8,power_law_1.2,0.8029325103759766
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,32,8,power_law_1.2,0.8034444427490234
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,32,8,power_law_1.2,0.8122073364257811
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,32,8,power_law_1.2,0.8323007965087891
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,32,8,power_law_1.2,0.9100236511230468
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,32,8,power_law_1.2,0.9726131439208985
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,32,8,power_law_1.01,8.813332519531249
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,32,8,power_law_1.2,1.1173426818847656
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,32,8,power_law_1.2,1.4386189270019532
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,32,8,power_law_1.2,1.6527066040039062
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,32,8,power_law_1.2,3.769890441894531
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,32,8,power_law_1.2,2.158228454589844
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,32,8,power_law_1.2,0.7760614776611329
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,32,8,power_law_1.2,2.9025140380859376
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,32,8,power_law_1.2,3.931717224121094
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,32,8,power_law_1.2,5.271882324218749
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,32,8,power_law_1.2,7.530068359374999
SGLang,0.5.10,NVIDIA H200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,32,8,power_law_1.2,10.3427197265625
